Incidence of infections in patients with giant cell arteritis: A cohort study




Giant cell arteritis (GCA) is the most frequent form of vasculitis in adults. We sought to estimate the infectious risk associated with GCA and its treatment.


We conducted a matched historical cohort study using data from The Health Improvement Research Network. Patients with newly diagnosed GCA were matched with up to 6 non-GCA patients by age, sex, general practice, and date of entry into the cohort. Random-effects Poisson regression models were used to obtain incidence rates and rate ratios for lower respiratory tract infections (LRTI), urinary tract infections (UTI), and sepsis, as well as for the subset of these that comprised serious infections (pneumonias, upper UTI, and sepsis). Effect modification by age, sex, and time since diagnosis of GCA was assessed.


A total of 1,664 patients with GCA were matched to 8,078 patients without GCA. Overall, 805 (48%) of the GCA patients and 3,007 (37%) of the non-GCA patients experienced ≥1 episode of systemic infection during followup, with adjusted rate ratios for LRTI, UTI, and serious infections of 1.48 (95% confidence interval [95% CI] 1.34–1.65), 1.27 (95% CI 1.10–1.46), and 1.55 (95% CI 1.22–1.96), respectively (P < 0.001 for all). The rate ratio for sepsis was 1.63 (95% CI 0.78–3.40, P = 0.20). Rate ratios for infection were highest in the first 6 months following diagnosis of GCA and in patients age <75 years, but did not vary by sex.


This is the first study to show that patients with GCA are at increased risk of systemic infections, particularly in the first few months following diagnosis. New GCA medications that allow steroid sparing are needed to treat this condition.


Giant cell arteritis (GCA) is the most frequent form of vasculitis in adults (1, 2). Incidence is generally higher in Northern regions (1). Reports of incidence (in cases per million person-years in population age >50 years) vary widely; for example, 328 in Norway (3), 101 in North-Western Spain (1), 25–46 in Germany (4), 11 in Turkey (5), and 3.4 in Hong Kong (6). GCA causes significant morbidity through visual loss (8–29% of patients), thoracic aorta aneurysm and dissection (7), and increased risk of stroke and coronary artery disease (1, 8–11). It is more frequent among women (ratio ∼3:1), and median age of disease onset is ∼75 years (2).

Glucocorticoids are the mainstay of therapy and are started at high doses and gradually tapered. One study reported total median prednisone doses of 6.5 gm (12); another reported that mean duration of treatment was 44.6 months (13). Another group reported that 83% of patients remained on steroids at 1 year (14). In most cases, treatment results in resolution of GCA and patients can be completely weaned off steroids, but 26.2% (15) to 62.4% (16) experience relapses requiring transitory increase in doses. Research to find alternative therapies or steroid-sparing regimens for patients with GCA has not yielded convincing results, with limited evidence of benefit using methotrexate or infliximab. The potential roles of adalimumab and abatacept are currently being evaluated (17), and anti–interleukin-6 tocilizumab seems promising in case reports but has not yet been assessed in a randomized study (18, 19). There are, therefore, unmet needs for steroid-sparing regimens in the treatment of GCA.

Steroids are known to increase susceptibility to infection via multiple pathways. They induce thinning and delayed healing of skin and mucous membranes, decrease migration, phagocytosis, and intracellular killing in macrophages, and suppress T cell–mediated cellular immunity (20). In a meta-analysis of 71 trials focusing on infectious risk linked to steroid use, infections occurred in 12.7% of patients randomized to steroids versus 8.0% of patients randomized to placebo, representing a risk ratio in steroid-treated patients of 1.6 (95% confidence interval [95% CI] 1.3–1.9) (21). However, the mean age of patients included in the individual studies ranged from 25.8–74.5 years, there was appreciable between-study variation in the doses, duration, administration routes, and indications for steroids, and only 4% (167 of 4,198) of patients had a rheumatologic condition. The burden of infection attributable to the use of high doses of steroids in GCA patients has not been well described in the literature. Case series have reported infection risks ranging from 15.8–31% (12, 13, 22, 23), but none had infections as their primary outcomes or compared the frequency of infection among GCA patients with that among a similar population without GCA. One controlled US study of 204 cases of GCA reported no increase in the incidence of herpes zoster, but did not include data about other infections (24).

The study of infectious risk for GCA patients is important, given that it is a frequent vasculitis, affects an older population, and is treated with considerable doses of steroids. In patients with GCA, both dysfunction of the immune system inherent to the disease and steroid treatment could lead to an increased risk of systemic infections. We therefore sought to compare the incidence of lower respiratory tract infections (LRTI), urinary tract infections (UTI), and sepsis in a cohort of patients with GCA to that in a cohort of same age and sex without GCA.

Significance & Innovations

  • This is the first study with a control population to quantify the increased risk of systemic infections among patients with giant cell arteritis (GCA).

  • Infections risks are increased by 27–55% depending on the type of infection, and increased risks are particularly prominent in the first 2 years after diagnosis.

  • GCA is associated with a greater increase in infection risk in younger patients.

  • There are unmet needs for new GCA medications that allow steroid sparing.


Study population.

This study was conducted on data from The Health Improvement Network (THIN) database, which contains anonymized patient data from more than 300 general practices throughout the UK, including medical diagnoses (coded using Read codes), medications, feedback from hospital referrals, and other health determinants collected by general practitioners (25). The population of patients contributing to THIN has been shown to be representative of the UK population in several respects (26), and studies have demonstrated the validity of diagnoses in THIN (27). The study population was a subset of a larger study on autoimmune diseases, which included patients registered with a general practice contributing data to THIN between January 1987 and February 2007; complete data were available for patients with any autoimmune disease (from which patients with GCA were selected) and patients without any autoimmune diseases that were matched by age, sex, general practice, and time of followup to the patients with autoimmune diseases (from which patients without GCA were selected).

Exposed cohort.

Patients were considered exposed to GCA if they fulfilled all of the following inclusion criteria. First, the patients had to have received a first diagnosis for GCA while they were being followed up at a contributing general practice. GCA diagnoses were ascertained using selected Read codes (see Supplementary Appendix A, available in the online version of this article at Patients with a past history of GCA were excluded, as were those with a first GCA diagnostic code that occurred within the first 6 months after joining a general practice (it has been shown that diagnoses occurring during this time window are often past diagnoses being added to the patient's file) (28). Second, the patients had to be age >40 years at the time of GCA diagnosis. Third, the patients had to have received a steroid prescription within 6 months of their first GCA diagnosis and a second within 6 months of the first. This case definition for GCA was used as it has been validated in a study based in the General Practice Research Database (a primary care database that contains patient data that overlap with the data in THIN) (29).

Cases of GCA were excluded from the cohort if they received a diagnosis for another autoimmune disease. This exclusion criterion was added to ensure consistency with the comparison group (patients without GCA) who came from the cohort of patients with no evidence of any autoimmune diseases.

Unexposed cohort.

For each GCA patient, up to 6 non-GCA patients were selected, matched for age (within 9 years), sex, and general practice, and registered with the practice at the time their matched GCA case was diagnosed. They also had to have at least 1 consultation or prescription in the 6 months before or in the year after their enrollment to avoid the inclusion of unexposed patients who had left the practice but had remained registered.

Date and followup.

Each patient started followup on their index date. For GCA patients, this was defined as the date of GCA diagnosis. For patients without GCA, the index date was that of their matched GCA-exposed counterpart. End of followup was defined as the earliest of end of registration with the general practice, death, end of contribution of data to THIN by the general practice, or February 2007.


Systemic infections were defined using Read codes and categorized in 4 groups. LRTIs included all diagnoses for pneumonias, acute bronchitis, or unspecified chest infections. UTIs included diagnoses of both upper and lower urinary tract infections. Sepsis represented all cases of sepsis, septic shock, or septicemia. From these 3 groups we generated a fourth category, serious infections, which included the subset of LRTI that specifically identified pneumonias, the subset of UTI that identified upper UTI, and all sepsis episodes. Recurrent events were included in the analysis. Diagnoses recorded less than 28 days apart were considered part of the same episode (consequently, the 28 days following each event were removed from followup time).

Definitions for covariates.

Predefined confounding variables were the following: sex, age, body mass index (kg/m2), smoking status, alcohol intake, heart failure, history of myocardial infarction, history of stroke, diabetes mellitus, chronic pulmonary obstructive disease (COPD), use of methotrexate, use of cytotoxic drugs (including chemotherapy drugs), influenza vaccination, and pneumococcal vaccination. All covariates were defined using therapy or Read codes.

Covariates that could vary over time were coded in a time-updated manner. For pneumococcal vaccination, it was assumed that 1 vaccination granted lifelong immunity. For influenza vaccination, followup time was divided into 1-year periods starting September 1 of each year, and patients were exposed to vaccination for that year if they were vaccinated during the flu vaccination campaign (between September 1 and March 31). For body mass index, smoking, and drinking habits, if no value was available before or at the start of followup, the first available value during followup was assigned to baseline. For values missing throughout followup, the missing indicator method was used.

A priori effect modifiers were sex, age (divided in quartiles of the age distribution), and time since diagnosis of GCA (divided into <6 months, 6 months to a year, second year, and third year following diagnosis).

Statistical analysis.

To assess differences in baseline characteristics, we used t-tests for difference in means, nonparametric K-sample tests for difference in medians, and chi-square tests for categorical variables. To model infection rates and rate ratios of infections, we used multivariable Poisson regression models with random effects to account for inclusion of recurrent events within the same patients.

For the outcomes of LRTI, UTI, and serious infections, all potential confounders were introduced and kept in the model. For the outcome of sepsis, due to the limited number of events, only covariates that were associated (P < 0.20) with sepsis in the univariable analysis were entered into the model and kept if they modified the point estimate of the effect of GCA on sepsis by >5%. Effect modification was modeled by introducing interaction terms in the models.

Ethics approval.

Ethics approval for this project was obtained by the London School of Hygiene and Tropical Medicine Ethics Committee and from the South Thames Multi-Centre Research Ethics Committee.


Figure 1 details the selection of patients in the study. Of the 5,241 GCA patients identified initially, 1,664 filled the inclusion criteria and were included in the study. These patients had 8,078 matched patients without GCA.

Figure 1.

Flow chart of patients included in study. GCA = giant cell arteritis; THIN = The Health Improvement Network.

Baseline characteristics of the study population are presented in Table 1. Median age was 71.5 years, with GCA patients slightly older than non-GCA patients. Sixty-nine percent of the population were women. Length of followup was similar in the 2 groups, and a total of 43,507 patient-years of followup were included in the study.

Table 1. Characteristics of patients with and without GCA*
CharacteristicGCA cohort (n = 1,664)Non-GCA cohort (n = 8,078)P
  • *

    GCA = giant cell arteritis; IQR = interquartile range; COPD = chronic obstructive pulmonary disease.

  • By chi-square test unless indicated otherwise.

  • Obtained using nonparametric K-sample test.

  • §

    Quartiles defined in the overall cohort.

Age at entry, median (IQR) years74.57 (67.75–80.12)72.88 (64.94–79.80)< 0.001
Female, no. (%)1,158 (70)5,519 (68)0.31
Smoking, no. (%)  < 0.001
 Nonsmoker833 (50)4,148 (51) 
 Ex-smoker204 (12)939 (12) 
 Current smoker503 (30)2,136 (26) 
 Unknown124 (7)855 (11) 
Body mass index (kg/m2), no. (%)  0.09
 <2095 (6)389 (5) 
 20–291,004 (60)4,798 (59) 
 ≥30219 (13)1,053 (13) 
 Unknown346 (21)1,838 (23) 
Drinking habits, no. (%)  0.49
 Non- or moderate drinker1,272 (76)6,042 (75) 
 Ex-drinker12 (0.7)74 (0.9) 
 Alcohol abuse9 (0.5)59 (0.7) 
 Unknown371 (22)1,903 (24) 
Length of followup, median (IQR) years3.59 (1.51–6.72)3.41 (1.49–6.39)0.17
 Total person-years7,61935,888 
Mean consultations per year, no. (%)  < 0.001
 First quartile§178 (11)2,126 (26) 
 Second quartile315 (19)2,059 (25) 
 Third quartile469 (28)2,023 (25) 
 Fourth quartile702 (42)1,870 (23) 
Type 2 diabetes mellitus, no. (%)195 (12)740 (9)0.001
Heart failure, no. (%)256 (15)808 (10)< 0.001
History of myocardial infarction, no. (%)164 (10)588 (7)< 0.001
History of stroke, no. (%)312 (19)1,159 (14)< 0.001
COPD, no. (%)227 (14)719 (9)< 0.001
Use of methotrexate, no. (%)39 (2)7 (0.1)< 0.001
Use of cytotoxic drugs, no. (%)347 (21)890 (11)< 0.001
Pneumococcal vaccination, no. (%)914 (55)3,786 (47)< 0.001
Influenza vaccination, no. (%)1,298 (78)5,524 (68)< 0.001

At baseline and during followup, patients with GCA were more likely to have experienced or to develop diabetes mellitus, heart failure, myocardial infarction, stroke, and COPD (Table 1). They were also more likely to receive treatments with methotrexate and cytotoxic drugs and to receive pneumococcal or influenza vaccination.

Figure 2 shows the Kaplan-Meier curves of time to first event for each systemic infection, according to GCA status. In total, 3,812 (39%) patients experienced at least 1 episode of systemic infection during followup (805 [48%] patients with GCA and 3,007 [37%] patients without GCA). LRTI, UTI, sepsis, and serious infections were present, respectively, in 618 (37%), 401 (24%), 14 (0.8%), and 126 (8%) of GCA patients and in 2,148 (27%), 1,492 (18%), 32 (0.4%), and 374 (5%) of non-GCA patients.

Figure 2.

Kaplan-Meier curves of time to first event for each outcome. LRTI = lower respiratory tract infection; GCA = giant cell arteritis; UTI = urinary tract infection.

Table 2 shows the results of the univariable and multivariable analyses for each type of infection, taking into account multiple events per patient. After adjusting for all covariates, GCA was associated with an increased risk for LRTI, UTI, and serious infections with rate ratios of 1.48 (95% CI 1.34–1.65), 1.27 (95% CI 1.10–1.46), and 1.55 (95% CI 1.22–1.96), respectively (P < 0.001 for all). There was only weak evidence of increased sepsis risk, with a rate ratio of 1.63 (95% CI 0.78–3.40, P = 0.20).

Table 2. Univariable and multivariable association between GCA and incidence of infection in 1,664 GCA patients and 8,078 non-GCA patients*
Exposure statusCrude modelAdjusted model
Events/person-yearsRate per 100 person-years (95% CI)RR (95% CI)RR (95% CI)
  • *

    GCA = giant cell arteritis; 95% CI = 95% confidence interval; RR = rate ratio; LRTI = lower respiratory tract infection; UTI = urinary tract infection.

  • Fully adjusted model for all, except sepsis model, includes age, smoking, alcohol use, body mass index, diabetes mellitus, heart failure, history of stroke, history of myocardial infarction, chronic obstructive pulmonary disease, influenza vaccination, pneumococcal vaccination, and use of cytotoxic drugs or methotrexate.

  • Due to the limited number of sepsis events, only covariates that changed the point estimate by ≥5% were kept in the model. Covariates for this analysis were entered in order of the magnitude of their association with the outcome in the univariable model. The adjusted model included age (modeled linearly), heart failure, and use of methotrexate.

  • §

    Serious infections included pneumonias, upper UTI, and all sepsis episodes.

 P  < 0.001< 0.001
 GCA1,534/76,13322.55 (20.40–24.93)1.67 (1.50–1.87)1.48 (1.34–1.65)
 Non-GCA4,434/35,876713.48 (12.83–14.17)1.01.0
 P  < 0.001< 0.001
 GCA954/76,15112.75 (11.23–14.48)1.34 (1.17–1.55)1.27 (1.10–1.46)
 Non-GCA3,250/358,8479.49 (8.93–10.09)1.01.0
 P  0.030.20
 GCA14/76,1570.23 (0.11–0.47)2.31 (1.08–4.93)1.63 (0.78–3.40)
 Non-GCA33/358,8740.10 (0.07–0.15)1.01.0
Serious infection§    
 P  < 0.001< 0.001
 GCA141/76,1422.27 (1.82–2.82)1.75 (1.38–2.24)1.55 (1.22–1.96)
 Non-GCA405/358,8091.29 (1.14–1.46)1.01.0

Table 3 shows the results of the analyses of the effect of GCA on infections by time since GCA diagnosis. For all outcomes except UTI, rate ratios for infections were highest within the first 6 months of followup and then declined over time.

Table 3. Variation in the effect of GCA on risk of infection by time since diagnosis in 1,664 GCA patients and 8,078 non-GCA patients*
Exposure statusTime since GCA diagnosis 
0–5 months6–12 monthsSecond yearThird yearP for interaction
  • *

    Values are the rate ratio (95% confidence interval) unless indicated otherwise. GCA = giant cell arteritis; LRTI = lower respiratory tract infection; UTI = urinary tract infection.

  • The interaction with followup time for sepsis events was modeled linearly to favor parsimony and because there was no evidence of departure from linearity (P = 0.99). Thus, the rate ratios presented for each time interval are those predicted by the model at the midpoint of the interval.

  • Serious infections included pneumonias, upper UTI, and all sepsis episodes.

 GCA1.72 (1.43–2.07)1.70 (1.40–2.06)1.53 (1.29–1.81)1.28 (1.06–1.53)0.031
 GCA1.44 (1.13–1.84)1.40 (1.07–1.82)1.30 (1.03–1.62)1.12 (0.88–1.43)0.49
 GCA3.39 (1.33–8.65)2.50 (1.15–5.47)1.59 (0.78–3.26)0.87 (0.33–2.32)0.034
Serious infection     
 GCA3.57 (2.21–5.76)1.79 (1.05–3.06)1.71 (1.01–2.87)1.07 (0.57–1.99)0.0015

Table 4 shows the effect of GCA by age. For all 4 outcomes, the highest rate ratio was seen in the youngest age category (<70 years) and then tended to decrease. There was no evidence of effect modification by age for sepsis and only weak evidence of such an effect for UTI. There was no evidence of effect modification by sex (data not shown). Taking into account the European League Against Rheumatism age criterion for diagnosis of GCA, we performed a sensitivity analysis in which we excluded the 21 GCA patients age <50 years at diagnosis (median age 47 years) and their matched non-GCA patients. Results, including effect modification by age, were unchanged (data not shown).

Table 4. Variation in the effect of GCA on risk of infection by age in 1,664 GCA patients and 8,078 non-GCA patients*
Exposure statusAge range 
<70 years70–74 years75–79 years>80 yearsP for interaction
  • *

    Values are the rate ratio (95% confidence interval) unless indicated otherwise. GCA = giant cell arteritis; LRTI = lower respiratory tract infection; UTI = urinary tract infection.

  • The interaction with age for sepsis events was modeled linearly in order to favor parsimony; there was no evidence of departure from linearity (P = 0.26). Therefore, the rate ratios presented for each time interval are those predicted by the model at ages 65, 72.5, 77.5, and 85 years (ages 65 and 85 years are close to the median age of patients included in the first and last categories).

  • Serious infections included pneumonias, upper UTI, and all sepsis episodes.

 GCA2.23 (1.86–2.66)1.44 (1.20–1.72)1.19 (1.01–1.41)1.28 (1.10–1.50)< 0.001
 GCA1.61 (1.25–2.06)1.11 (0.87–1.41)1.19 (0.96–1.48)1.22 (1.00–1.48)0.097
 GCA2.77 (0.77–9.96)2.07 (0.87–4.90)1.70 (0.82–3.55)1.27 (0.51–3.16)0.34
Serious infection     
 GCA3.30 (1.85–5.90)2.05 (1.16–3.63)1.41 (0.84–2.35)1.22 (0.89–1.66)0.024


In this population-based retrospective cohort study, we found that risk of systemic infection in patients diagnosed and treated for GCA was higher than in an age- and sex-matched population. There was strong evidence that LRTI, UTI, and serious infections were increased in patients with GCA, with increased risks between 27% and 55%. There was only weak evidence of increased risk of sepsis, but the number of patients experiencing sepsis was small and power was limited.

The proportion of GCA patients experiencing at least 1 infectious episode in our study (48%) was higher than the 15.9% and 30.8% reported previously in smaller numbers of GCA patients from France (14) and the US (12). This could be due in part to a less inclusive definition of infectious events in these studies. The magnitude of effects of GCA on LRTI and serious infections was similar to a point estimate of 1.6 found in a meta-analysis comparing patients randomized to steroids or placebo (21).

We found that the effect of GCA on incidence of each systemic infection (except UTI) decreased with increasing time, with the highest rate ratios for infection occurring in the first 6 months after GCA diagnosis. This is consistent with the highest infection risk occurring while GCA patients are on high doses of steroids. We found no evidence for effect modification by time since the index date for the outcome of UTI. It is possible that immunosuppression by steroids has less effect on urinary pathogens (mainly Escherichia coli) than on respiratory pathogens (e.g., Streptococcus pneumoniae, Haemophilus influenzae) (20). Finally, when observing infections in GCA versus non-GCA patients, the rates mostly converged in the second to third year after the index date, a period corresponding to the duration of GCA and its treatment (13, 22).

Interestingly, we found that GCA was associated with a greater increase in risk of infection in younger patients (age <70 years). This was unexpected, and possible reasons for this finding include: 1) younger patients might have a more active immune system and therefore have more aggressive disease requiring more immunosuppressive medication, 2) GCA and its treatment might speed up the immunosenescence process in younger individuals, making them more comparable to older individuals, and 3) there could be some survivor bias. Older individuals made it to an older age because they had more robust immune systems. This finding remains to be confirmed in other cohorts and for other autoimmune diseases.

Strengths of this study include its large size compared to previous studies and the presence of a comparison group without GCA selected from the same population as the GCA cases. We had very good power to detect increases in infectious risk (>90% for a rate ratio of 1.5) for all subtypes of systemic infections except sepsis. The matching by age and sex enabled comparability of the 2 groups, matching by general practice allowed partial control of hard to measure confounders such as prescription and diagnostic habits, as well as potential geographic variations in incidence of infections, and matching by index date allowed control of secular trends in treatment of GCA and diagnosis of infections, as well as seasonality in incidence of infections. The time-updated coding of covariates that varied over time is also a strength of our study, allowing more precise adjustment throughout long followup periods.

Some limitations are inherent to the design of the study and data source. The use of administrative data and reliance on diagnostic codes could have resulted in some degree of misclassification of GCA status, covariates, and systemic infections. As diagnosis of GCA was prior to assessment of the outcomes of interest, any misclassification of GCA status could not be differential and would only bias results toward the null. We used a validated definition of GCA that has been shown to have a high predictive positive value of 91% (95% CI 79–98) (29).

Misclassification of systemic infections could also have occurred. It is possible that this misclassification could be differential with respect to GCA status, as patients followed up for GCA were seen more frequently by their general practitioners than non-GCA patients, giving them more opportunities to be diagnosed with systemic infections (ascertainment bias), and general practitioners might also be more prone to diagnose an infection in patients receiving steroids. On the other hand, steroids can mask the symptoms of infections and therefore make them harder to diagnose. Differential misclassification could bias the results away from the null and toward a greater infectious risk associated with GCA. We did not adjust for consultation rates because that would adjust for disease severity, which was part of our exposure of interest. However, ascertainment bias is unlikely to have occurred for the more serious outcomes of sepsis and serious infections, and the consistency of the effect seen with these 2 outcomes is reassuring.

The selection process in our study also introduces some limitations to the generalizability of our findings. The GCA and non-GCA patients were selected from a subset of a THIN data set comprising individuals with autoimmune diseases and matched individuals without any autoimmune diseases. As all non-GCA individuals were unexposed to any other autoimmune condition, it was necessary to exclude from the study the GCA patients who also had another autoimmune disease. Patients who had a nonspecific code for diabetes mellitus and use of insulin were also excluded from the original study because of concerns that these patients had type 1 (autoimmune) diabetes mellitus. This will have led to the exclusion of some patients with insulin-dependent type 2 diabetes mellitus. GCA and its treatment have been associated with incidence of diabetes mellitus (30–33), and diabetes mellitus increases infectious risk (34), so the risk of infection associated with GCA in a population including such individuals could be even higher.

We could not assess dose-response analyses with the steroid doses given, as data on steroid doses prescribed were scarce. The analysis of effect modification by time since diagnosis of GCA was used as a proxy and is an indirect reflection of the impact of initially high and then decreasing steroid doses.

In conclusion, this is the first study to demonstrate an increased infectious risk for systemic infections in patients with GCA compared to patients of the same age and sex. This increased risk is likely to be due to both the disease itself and to treatment with steroids. Treating physicians should be aware of that increased risk, particularly at the beginning of treatment. The greater rate ratio of infection observed in younger patients is an interesting finding and deserves further investigation. New medications that allow steroid sparing for patients with GCA are needed, and their ability to reduce infectious risk must be prospectively assessed in randomized trials given that infections are a frequent and important comorbidity for patients with GCA.


Both authors were involved in drafting the article and revising it critically for important intellectual content, and both authors approved the final version to be submitted for publication. Dr. Durand had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Durand, Thomas.

Acquisition of data. Durand, Thomas.

Analysis and interpretation of data. Durand, Thomas.


The authors would like to acknowledge the help of Katherine Fielding, Kathy Baisley, and Ian Douglas for assistance with programming and data management. They also thank Dr. Yves Troyanov for his help in reviewing the manuscript. Dr. Durand is grateful to all members of the Internal Medicine Service of the Centre Hospitalier de l'Universtité de Montréal.