Increased risk of thoracic aortic complications among patients with giant cell arteritis: a nationwide, population-based cohort study

Abstract

Objective

To assess the risk of aortic aneurysms (AA), aortic dissections (AD) and peripheral arterial disease (PAD) among patients with GCA.

Methods

In this nationwide, population-based cohort study using Danish national health registries, we identified all incident GCA patients ≥50 years between 1996 and 2018 who redeemed three or more prescriptions for prednisolone. Index date was the date of redeeming the third prednisolone prescription. Case definition robustness was checked through sensitivity analysis. We included general population referents matched 1:10 by age, sex and calendar time. Using a pseudo-observation approach, we calculated 5-, 10- and 15-year cumulative incidence proportions (CIP) and relative risks (RR) of AA, AD and PAD with death as a competing risk.

Results

We included 9908 GCA patients and 98 204 referents. The 15-year CIP of thoracic AA, abdominal AA, AD and PAD in the GCA cohort were 1.9% (95% CI 1.5, 2.2), 1.8% (1.4–2.2), 1.0% (0.7–1.2) and 4.8% (4.2–5.3). Compared with the referents, the 15-year RR were 11.2 (7.41–16.9) for thoracic AA, 6.86 (4.13–11.4) for AD, 1.04 (0.83–1.32) for abdominal AA and 1.53 (1.35–1.74) for PAD. Among GCA patients, female sex, age below 70 years and positive temporal artery findings were risk factors for developing thoracic AA. The median time to thoracic AA was 7.5 years (interquartile range 4.4–11.2) with a number needed to be screened of 250 (167–333), 91 (71–111) and 53 (45–67) after 5, 10 and 15 years.

Conclusion

Patients with GCA have a markedly increased risk of developing thoracic AA and AD, but no increased risk of abdominal AA.

Introduction

GCA is a large- and medium-sized vessel vasculitis [1]. The vascular inflammation may disrupt elastin and collagen fibres in the vessel wall, which causes a mechanical weakness, leading to an increased risk of aortic aneurysms (AA) and aortic dissections (AD) [2). Still, the absolute risk of AA and AD among patients with GCA remains unclear and estimates vary significantly between studies. AA has been reported to develop in 2–20% of patients with GCA and thoracic AA, being up to 17 times more frequent compared with the general population [3–6]. A meta-analysis estimated a 3-fold increased risk of AA among patients with GCA as compared with the general population [2]. Also, the vessel inflammation can cause stenosis, increasing the risk of peripheral arterial disease (PAD) [7]. All these vascular complications are associated with increased morbidity and mortality [8–10]. Thus, the aim of this study was to assess the risk of AA, AD and PAD among patients with GCA in a nationwide, population-based study and to compare these risks with those in the general population.

Methods

Study design

This is a Danish nationwide, population-based cohort study assessing the risk of AA, AD and PAD among patients with GCA as compared with the general population.

Setting

Denmark has a population of ∼5.8 million. Access to healthcare is free of charge and paid for through taxes [11]. At birth or immigration, all Danish residents receive a unique civil registration (CPR) number, allowing linkage of data across nationwide medical and administrative registries at the individual level [12].

GCA cohort

We identified all individuals aged 50 years or older with their first recorded primary or secondary diagnosis of GCA in the Danish National Patient Registry (DNPR) between 1 January 1996 and 31 December 2018. The DNPR contains information on all somatic hospital admissions since 1977 and all hospital out-patient visits since 1995. The International Classification of Diseases system (ICD-10) has been used to classify diagnoses since 1994 [13]. In Denmark, the majority of GCA patients are evaluated in secondary care and Danish national guidelines endorse referral of all patients suspected of GCA for rheumatologic evaluation. Diagnoses given by primary care physicians or private practice rheumatologist are not included in the DNPR. The positive predictive value (PPV) of GCA in the DNPR is 62% [14]. To increase the specificity of the GCA diagnosis, we identified all patients with GCA who had redeemed three or more prescriptions for prednisolone within 6 months after the date of the GCA diagnosis (Fig. 1), through the Danish National Prescription Registry (DPR). This definition has previously shown a PPV of 78% and a completeness of 91% for the GCA diagnosis. Among false-positive GCA diagnoses, the most common diagnosis was PMR [14]. The DPR contains information on all prescriptions redeemed at Danish pharmacies since 1994. Drugs are classified according to the Anatomical Therapeutic Chemical classification of drugs [15]. Index date was defined as the date of redeeming the third prednisolone prescription. Diagnostic codes are shown in supplementary Table S1, available at Rheumatology online.

Reference cohort

A reference cohort was sampled by individually matching every GCA patient on sex, age, and calendar time with up to 10 individuals from the general population without a history of GCA using the Danish Civil Registration System (CRS). The CRS holds information on CPR number, name, address, date of birth, vital status, civil status, citizenship and migration on all residents in Denmark [12]. The reference cohort was assigned with the same index date as their matching GCA patient.

Outcomes

Primary outcomes were thoracic and abdominal AA, AD and PAD. Secondary outcomes were aortic surgery, surgery for PAD, and minor or major amputation of upper- or lower extremities. PAD was defined as lower limb claudication due to the high positive predictive value of this diagnostic code, and to the fact that ICD-10 has no specific code for chronic upper extremity ischaemia. Minor amputations were defined as amputations at the wrist/ankle. We obtained information on all primary outcomes through the DNPR, both prior to the index date and during follow-up. Information on surgery for PAD was obtained through the Danish Vascular Registry (KARBASE). This registry contains information on all patients undergoing vascular interventions (surgical and endovascular) at any vascular surgical department in Denmark since 1996 [16]. From the CRS we obtained information on the date of death or migration. Diagnostic codes for outcome variables are shown in supplementary Table S1, available at Rheumatology online.

Covariates

We obtained information on redeemed prescriptions for prednisolone, antithrombotics, antihypertensives, lipid-lowering drugs and antidiabetics from 1 year before the index date and during follow-up through the DPR. Low-dose aspirin is included under antithrombotics. Although low-dose aspirin is available over-the-counter, the potential for identification of individual-level use of low-dose aspirin from prescriptions registries in Denmark is high, with 62% increasing to 92% of sales in the period 1999–2012 being prescribed on an individual level [17]. Anatomical Therapeutic Chemical codes are shown in supplementary Table S1, available at Rheumatology online. Furthermore, we identified temporal artery biopsies (TABs) performed within ±3 months of the GCA diagnosis using the Danish Pathology Registry. This registry contains information on every pathology specimen analysed in Denmark since 1997 and incomplete records from 1970 to 1997. Specimens are classified according to the Danish version of Systemized Nomenclature of Medicine (SNOMED) [18]. We defined a positive TAB as a TAB specimen showing inflammation. SNOMED codes are shown in supplementary Table S1, available at Rheumatology online. Also, we obtained information on the highest measured level of CRP and ESR within ±1 month of the index date through the Registry of Laboratory Results for Research. This registry contains information on analysed blood samples taken in any public or private hospital in Denmark since 2013. Blood samples are coded according to the international NPU (Nomenclature, Properties and Units) coding system [19]. NPU codes are shown in supplementary Table S1, available at Rheumatology online. Lastly, we obtained information on all comorbid diseases up to 10 years prior to the index date and calculated the Charlson’s Comorbidity Index based on all diseases diagnosed in this period (supplementary Table S2, available at Rheumatology online) [20].

Statistics

The GCA and reference cohorts were considered at risk from the time of the index date and followed until outcome, migration, death or end of study (31 December 2018), whichever came first. Outcomes were analysed separately and participants with a history of the outcome of interest prior to the index date were excluded.

Time-to-event analyses were performed using a pseudo-observation approach [21]. Pseudo-observations were generated at 5, 10 and 15 years following the index date with death as a competing risk factor. They were used in a generalized linear model with a robust estimator of variance to estimate unadjusted and adjusted cumulative incidence proportions (CIPs), risk differences (RDs) and relative risks (RRs). Age (continuous variable), sex and Charlson’s Comorbidity Index (categorized as 0, 1 and ≥2) [22] were used as variables in the adjusted analysis. We generated Kaplan–Meier plots stratified by calendar time to ensure that the assumption of independent censoring was not violated. Further, we plotted the proportion of covariates over calendar time to ensure the assumption of independent covariates was not violated. Based on the CIPs of AAs we calculated the number of GCA patients that needed to be screened. To take immortal time bias into consideration, we performed a sensitivity analysis including all incident cases of GCA without conditioning on redeemed prednisolone prescriptions. Here, index date was defined as the date of the GCA diagnosis. In another sensitivity analysis, we analysed all incident cases of GCA with a positive TAB only. Data were analysed using StataCorp 2019 (Stata Statistical Software: Release 16, StataCorp LLC, College Station, TX, USA).

Ethical approval

According to Danish law, the study was approved by the Danish Data Protection Agency (1-16-02-939-17). Further ethical approval was not required according to Danish law.

Results

Characteristics

We included 9908 GCA cases and 98 204 reference individuals (supplementary Fig. S1, available at Rheumatology online). The GCA and reference cohort showed comparable baseline characteristics on age, sex, comorbidity and medication use prior to the index date (Table 1). Among GCA patients, 968/9908 (9.8%) were registered with chronic obstructive pulmonary disease at the time of diagnosis, and 7757/98 204 (7.8%) among reference individuals. This GCA cohort has been described previously [23, 24].

Primary outcomes

Aortic aneurysms

Among GCA patients, 118 developed thoracic AA and 115 developed abdominal AA during a median follow-up time of 6.0 years [interquartile range (IQR) 2.5–10.4]. For thoracic AA, the CIP was 0.4%, 1.1% and 1.9%, respectively, after 5, 10 and 15 years. Compared with the reference cohort, both the RD and the RR of thoracic AA among GCA patients increased over time, with a RD of 1.6% (1.2–2.0) and an RR of 11.2 (7.41–16.9) after 15 years (Table 2, Fig. 1). The median time to thoracic AA was 7.5 years (IQR 4.4–11.2). The risk of thoracic AA was highest among women, patients below 70 years of age and among patients with a positive TAB (Table 3). Sensitivity analysis, including all incident GCA cases with a positive TAB, showed an RR of thoracic AA of 23.0 (11.9–44.2) after 15 years (supplementary Table S3, available at Rheumatology online). Additional sensitivity analysis including all incident GCA cases without conditioning on redeemed prednisolone prescriptions did not alter the estimates significantly (supplementary Table S4, available at Rheumatology online). The number of GCA patients that needed to be screened to identify one case of thoracic AA after 5, 10 and 15 years was 250 (167–333), 91 (71–111) and 53 (45–67), respectively. For female TAB-positive GCA patients below 70 years of age, the number that needed to be screened was 31 (21–59) and 16 (11–26), respectively, after 10 and 15 years. The CIP of abdominal AA in the GCA cohort was 0.6%, 1.2% and 1.8%, respectively, after 5, 10 and 15 years. Compared with the reference cohort, GCA patients showed a similar risk of abdominal AA with a RD of 0.3% (−0.1 to 0.7) and a RR of 1.04 (0.83–1.32) after 15 years (Table 2, Fig. 1). At the time of diagnosis, a larger proportion of GCA patients received imaging, including CT, MR, US and angiography scans of the thorax and abdomen, compared with the reference cohort (supplementary Fig. S2, available at Rheumatology online). However, during follow-up, the increase in the number of patients receiving imaging was similar in the GCA and reference cohorts (supplementary Fig. S2, available at Rheumatology online). The median number of imaging procedures received per individual during follow-up was 3 (IQR 1–9) in the GCA cohort and 1 (IQR 1–4) in the reference cohort. When excluding imaging procedures performed within the first 3 months following the diagnosis, this number was 1 (IQR 0–6) in the GCA cohort and 1 (IQR 1–4) in the reference cohort. In the GCA cohort, 97/9908 (1%) were registered with a death related to AA or AD during follow-up, corresponding to 2.2% of all 4430 deaths. In the reference cohort, 387/98 204 (0.4%) were registered with a death related to AA or AD during follow-up, corresponding to 0.9% of all 41 971 deaths.

Aortic dissections

A total of 61 GCA patients developed an AD during follow-up. The CIP of AD in the GCA cohort was 0.2%, 0.6% and 1.0%, respectively, after 5, 10 and 15 years. After 15 years, the RD and RR of developing AD for GCA patients were 0.8% (0.5–1.1) and 6.86 (4.86–11.4), respectively (Table 2, Fig. 1). The median time to an AD diagnosis was 7.6 years (IQR 3.8–11.8). Of the 61 patients with AD, 39 (64%) had an AA diagnosis as well. In 18 (46%) of the patients, AD occurred at the time of the AA diagnosis, and in 21 (54%) of the cases, AD occurred after the AA diagnosis [median 1.5 years (IQR 0.3–4.9)]. The risk of AD was higher among GCA patients aged below 70 years and among women (supplementary Fig. S3, available at Rheumatology online). The number of GCA patients that needed to be screened to identify one case of aortic dissection is 500 (333–1000), 167 (143–250) and 100 (83–143) after 5, 10 and 15 years following the diagnosis, respectively.

Peripheral arterial disease

A total of 336 GCA patients were diagnosed with PAD during follow-up. The CIP of PAD in the GCA cohort was 2.0%, 3.7% and 4.8% after 5, 10 and 15 years, respectively. After 15 years, the RD and RR of developing PAD for GCA patients were 1.7% (1.2–2.3) and 1.53 (1.35–1.74), respectively, compared with the reference cohort (Table 2 and Fig. 1).

Secondary outcomes

Aortic surgery

The CIP of thoracic aortic surgery in the GCA cohort was 0.3%, 0.6% and 1.0% after 5, 10 and 15 years, respectively. GCA patients had an increased risk of thoracic aortic surgery with a RR of 12.6 (7.52–20.9) after 15 years compared with the reference cohort (Table 4, Fig. 2). The risk of abdominal aortic surgery did not differ between the GCA and the reference cohort (Table 4, Fig. 2).

Surgery for PAD and extremity amputation

The CIP of PAD surgery in the GCA cohort was 0.4%, 0.8% and 1.0% after 5, 10 and 15 years, respectively. Compared with the reference cohort, GCA patients had an increased risk of PAD surgery with a RR of 1.45 (1.09–1.92) after 15 years and also an increased risk of minor and major amputations with RRs of 1.81 (1.35–2.44) and 1.54 (1.21–1.97), respectively (Table 4, Fig. 2).

Discussion

In this first, nationwide, population-based cohort study to assess the risk of AA, AD and PAD among patients with GCA, we found that patients with GCA had a markedly increased risk of developing thoracic AA, but not abdominal AA. This is consistent with GCA inflammation mainly affecting the thoracic aorta [25]. Among GCA patients, female sex, age below 70 years, and a positive TAB were associated with an increased risk of thoracic AA. Contrary to this, one population-based study has previously found an increased risk of AA among male GCA patients [26]. However, this included both abdominal and thoracic AA, and male sex is a well-known risk factor for abdominal AA [27].

Diagnostic codes for AA and AD have been validated in the DNPR and show a PPV of 100% and 92%, respectively [28]. However, the sensitivity of these diagnostic codes is unknown. AAs are often asymptomatic and discovered accidentally, and patients with GCA are not routinely screened for the development of AA [29]. Furthermore, the mortality following a ruptured AA is 85–90%, and about 50% die before they arrive at the hospital and are thus likely never diagnosed [30]. Also, 28% of AAs in the DNPR had an unspecified location and were not included in the location-specific analysis of AA. Consequently, we expect this register-based study to underestimate the incidence of AA among patients with GCA. Surveillance bias is another potential limitation, i.e. GCA patients being more likely to undergo imaging that could incidentally discover asymptomatic AAs. This could have made us overestimate the risk of thoracic AA among patients with GCA compared with the general population. However, GCA patients did not seem to undergo significantly more imaging during follow-up compared with the general population. Also, GCA patients were shown to have a markedly increased risk of undergoing thoracic but not abdominal aortic surgery, which emphasizes that GCA not only increases the risk of thoracic AA but also causes significant thoracic AAs requiring surgical intervention. Additional limitations include that Danish national registries do not have information on important lifestyle factors such as smoking, exercise and weight. Therefore, we cannot adjust for these, which could potentially bias our estimates in both directions depending on the distribution in the two cohorts. Smaller follow-up cohort studies have reported incidence proportions of AA among GCA patients in the range of 2–20% [4, 6, 7, 31–34]. However, most of these studies did not account for death as a competing risk, which could have made them overestimate the risk of AA. Today, more GCA patients are diagnosed with primarily large vessel disease due to an increasing use of imaging of large vessels [35–37]. Large vessel inflammation has been shown to be a strong predictor of subsequent aortic dilation among patients with GCA [38]. Hence, higher rates of aortic aneurysms could be expected among patients with GCA in the future.

The treatment of choice in GCA remains glucocorticoids (GCs) [39]. GCs are well-known for their severe side-effects, including hyperlipidaemia and hypertension, which have been linked to the risk of developing AAs and ADs. Hence, the GC treatment itself could play a role in the risk of developing AA and AD. It is not possible to separate the effect of disease and treatment in our study design, which could potentially have made us overestimate the risk of AA among patients with GCA. Immortal time bias in the main analysis is another potential limitation, since GCA patients were required to have survived until they had redeemed three prednisolone prescriptions. This might have made us underestimate the risk of AA and AD, as GCA patients dying before this time were excluded from the analyses. However, sensitivity analysis including all incident GCA cases without making a condition of redeemed prednisolone prescriptions did not alter the estimates significantly. Therefore, we expect that immortal time bias had little impact on our study results.

Screening for the development of AA among patients with GCA is highly debated [2] and international guidelines do not currently recommend routine screening [40]. The risk of thoracic AA-related adverse events, such as rupture and death, increases with the size of the thoracic AA and surgical intervention is recommended when the thoracic AA reaches a diameter of 5.0–6.5 cm, dependent on sex and comorbidity [41]. The average diameter growth rate of a thoracic AA is ∼0.10 cm/year [41]; however, the growth rate of thoracic AAs among patients with GCA is largely unknown. Only one study, based on medical chart review of 1450 GCA patients of whom 435 had identified aortopathy, has reported an estimated annual aneurysmal growth rate of 0.15 cm/year [42], indicating a more rapid growth of GCA-related AAs. In our cohort, the median time to a thoracic AA diagnosis was 7.5 years, supporting that AAs develop more rapidly among patients with GCA. Also, we found that the risk of thoracic AA varied significantly between GCA subgroups, with female TAB-positive GCA patients below 70 years of age showing the highest risk. Although GCA substantially increased the relative risk of thoracic AA, the absolute risk remained low (<7%) due to the overall rarity of the disorder. Which GCA patients to consider for screening should be weighed against the severity of unrecognized thoracic AA and the potential side-effects and costs of screening [41]. We have previously published data on cause-specific mortality in GCA, showing an increased risk of deaths due to cardiovascular diseases [24]. Also, the proportion of deaths related to AA or AD were over twice as high in the GCA cohort as compared with the reference cohort. Further research is needed to determine benefits and the optimal time point for screening among patients with GCA.

Few studies have estimated the risk of PAD in GCA [7, 43]. Our finding of an increased risk of PAD in patients with GCA thus adds to the existing literature. The PPV of PAD diagnostic codes in the DNPR has been estimated at 91% [28]. Our data support that patients with GCA have an increased risk of undergoing surgery for PAD and both minor and major amputations of extremities. The increased risk of PAD might be due to GC-related side-effects; however, it is not possible to separate the effects of disease and treatment in this study. Statin treatment at the time of the GCA diagnosis might prevent or postpone the time to symptomatic PAD and peripheral revascularization or amputations [44]. However, this has not been specifically examined in patients with GCA.

In conclusion, despite the fact that the absolute risk remained low, patients with GCA have a markedly increased relative risk of developing thoracic aortic aneurysms, but no elevated risk for abdominal aortic aneurysms. Risk factors for developing thoracic aortic aneurysms were female sex, age below 70 years and a positive TAB. Our findings provide conservative risk estimates and risk profiles that may impact on the long-term monitoring program in GCA.

Acknowledgements

Data for this study was partly provided by the Danish Vascular Registry.

Funding: This study was supported by Aarhus University and the Danish Rheumatism Association (grant numbers: R163-A5661 and R155-A4747).

Disclosure statement: The authors have declared no conflicts of interest.

Data availability statement

Data not available. The Danish Health Data Authority does not allow the sharing of data from national registries.

Supplementary data

Supplementary data are available at Rheumatology online.

References