Patients with giant cell arteritis have a lower prevalence of diabetes mellitus: A systematic review and meta-analysis.

OBJECTIVE
The aim of this study is to compare the prevalence of diabetes mellitus (DM) in patients who are recently diagnosed with giant cell arteritis (GCA) with age- and sex-matched controls.


METHODS
We conducted a systematic review and meta-analysis of observational studies that (1) consisted of GCA cohort and non-GCA cohort that was randomly selected from the same population and (2) provided prevalences of DM at the time of diagnosis for patients with GCA and at the index date for controls. Pooled odds ratios and 95% confidence intervals (CIs) were calculated using a random-effect, Mantel-Haenszel analysis.


RESULTS
Five studies with 903 patients with GCA and 1064 controls were identified and included in our data analysis. We demonstrated a statistically significant lower prevalence of DM among patients with GCA with the pooled OR of 0.74 (95% CI, 0.57-0.97).


CONCLUSION
At diagnosis, patients with GCA had a lower prevalence of DM. Whether DM could be a protective factor against the development of GCA needs further investigations.


Introduction
Giant cell arteritis (GCA) is the most common form of vasculitis in Western countries and is characterized by medium-and largevessel granulomatous arteritis [1]. GCA typically affects adults older than 50 years of age with a peak incidence among those aged 75-85 years [2,3]. Women are more frequently affected than men [4]. Patients with GCA usually present with signs and symptoms of systemic inflammation and cranial ischemia including headache, jaw claudication, scalp tenderness, and visual changes [1,5]. GCA is associated with significant morbidity and mortality secondary to its ischemic complications, such as permanent visual loss from ischemic optic neuropathy, cerebrovascular accident and large-vessel stenosis, aneurysm and dissection [1].
A low prevalence of diabetes mellitus (DM) at the diagnosis of GCA has been observed in several cohorts [4,6,7]. Moreover, a recent study found an inverse relationship between risk of developing GCA and body mass index (BMI) [8], which is a strong risk factor for DM. This raises the question of whether patients with GCA have a lower prevalence of DM compared with a sex-and age-matched population. Nevertheless, results of epidemiological studies attempting to answer this question were inconclusive as three studies showed a lower prevalence of DM among patients with GCA but without reaching a statistical significance [9][10][11], one study showed a significantly lower prevalence of DM among patients with GCA [12] and one study showed an equal prevalence between the two groups [13]. Thus, to further investigate this association, we conducted a systematic review and meta-analysis of epidemiological studies that compared the prevalence of DM in patients with GCA versus non-GCA controls.

Search strategy
Two investigators (P.U. and S.U.) independently searched published studies indexed in MEDLINE and EMBASE database from inception to October 2014 using the search strategy described in Supplementary data. References of selected retrieved articles were also manually searched.

Inclusion criteria
The inclusion criteria were as follows: (1) observational study that consisted of two cohorts. One was a GCA cohort. The other one was non-GCA cohort that was randomly selected from the same population; (2) prevalences of DM at the time of diagnosis for patients with GCA and at the index date for controls were provided. Study eligibility was independently determined by each investigator noted above. Any disagreements were resolved by consensus with the third investigator (A.S.)

Data extraction
Three investigators (P.U., S.U., and A.S.) independently extracted data from each study using a standardized data collection form. This information was extracted: last name of the first author, title of the study, year of publication, country where the study was conducted, number of cases and controls, basic epidemiological data of the cohorts, criteria used for the diagnosis of GCA, and prevalence of DM. Any discrepancies in data extraction were resolved by consensus.

Statistical analysis
Mantel-Haenszel odds ratios (OR) with 95% confidence intervals (CI) were calculated and pooled in this meta-analysis using Review Manager 5.3 software from the Cochrane Collaboration (London, United Kingdom). In light of the high likelihood of between study variance with the different study designs and populations, we utilized a random-effect model rather than a fixedeffect model. Cochran's Q-test was used to determine the statistical heterogeneity of this study. This test was complemented with the I 2 statistic, which quantified the proportion of the total variation across studies that is due to heterogeneity rather than chance. A value of I 2 of 0-25% indicates negligible heterogeneity, 26-50% indicates low heterogeneity, 51-75% indicates moderate heterogeneity, and 76-100% indicate high heterogeneity [14]. Funnel and Egger's linear regression methods [15] were utilized for the evaluation of publication bias using Comprehensive Meta Analysis version 2.2 software (Biostat, Englewood, NJ).

Results
Our search strategy yielded 4112 potentially relevant studies (1058 articles from MEDLINE and 3054 articles from EMBASE). After the exclusion of 1002 duplications, 3110 articles underwent title and abstract review. Three thousand and fifty-three articles were excluded as they clearly did not have two cohorts or were not conducted in patients with GCA, leaving 57 articles full-length article review. Forty-six of them were excluded since they were descriptive studies without a control group or used patients with negative temporal artery biopsy as their control group. An article was excluded as it identified the diagnosis of DM through medication database, leading an incompleteness of case identification (i.e. cases of diet-controlled DM would be missed) [16]. One study was excluded because it excluded subjects with baseline cardiovascular disease from its cohort [17], while another study was excluded because it reported the incidence of DM after the diagnosis of GCA but did not report the prevalence at the index date [18]. Eight studies met our eligibility criteria [9][10][11][12][13][19][20][21]. However, the studies by Machado et al. [19], Schmidt et al. [12], and Udayakumar et al. [20] as well as the studies by Duhaut et al. [21] and Le Page et al. [10] utilized the same or overlapping  Pfadenhauer et al. [9] Le Page et al. [10] Gonzalez-Juanatey et al. [11] Schmidt et al. [12] Hafner et al. [ [20], and Duhaut et al. [21] as the data from the study by Schmidt et al. [12] and Le Page et al. [10] were more comprehensive, leaving five studies with 903 patients with GCA and 1064 controls for the data analyses. Figure 1 illustrates our search methodology and article review process. The detailed characteristics of the included studies are described in Table 1.
Most of the included studies demonstrated a lower prevalence of DM among patients with GCA though without enough power to reveal a statistical significance. Our meta-analysis, taking the advantage of combining all the existing data, was able to demonstrate a statistically significant lower prevalence of DM among patients with GCA with the pooled OR of 0.74 (95% CI, 0.57-0.97). The statistical heterogeneity was negligible with an I 2 of 0%. Figure 2 demonstrates the forest plots of our findings as well as risk of bias for each study.
To confirm the robustness of our results, we performed a sensitivity analysis by excluding the studies by Pfadenhauer et al. [9] as this study, unlike other studies, used subjects older than 65 years old as its control group instead of age-and sex-matched subjects. In fact, the percentage of female subjects was significantly lower in the control group (68.8 and 47.3% in case and control group, respectively), raising a concern over the comparability of the two groups. However, the exclusion of this study did not significantly alter the result as the pooled OR only slightly decreased to 0.72 (95% CI, 0.55-0.95).

Evaluation for publication bias
Funnel plot to evaluate publication bias is shown in Figure 3. The graph is fairly symmetric and, thus, do not provide a suggestive evidence for publication bias. Furthermore, there was no evidence of publication bias detected by Egger's regression test (p ¼ 0.8).

Discussion
Although previous cohort studies have suggested a lower baseline prevalence of DM among patients with GCA, our study is the first meta-analysis to demonstrate a statistically significant lower prevalence of DM compared with sex-and age-matched controls with a pooled OR of 0.74. The validity of our results was confirmed by sensitivity analysis and a low statistical heterogeneity. It is unclear as to why patients with GCA have a lower prevalence of DM. It could be a result of confounders as DM is a major lifestyle-related disease, which might be associated with nutrition status, exercise, alcohol use, and smoking. Another possible hypothesis is that DM could be a protective factor against the development of GCA.
Giant cell arteritis is an antigen-driven inflammatory process, although the inciting antigen(s) have yet to be identified [22]. Molecular studies of large-vessel vasculitis have suggested that the inciting antigen(s) are recognized by dendritic cells residing in the vessel wall, which then initiate the pathogenic cascade by recruiting T cells and macrophages to form granulomas [23][24][25]. Patients with DM may have a decreased responsiveness of T cells to the putative antigen(s) and, thus, may have a lower likelihood of developing GCA. The decreased responsiveness of T cells of diabetic patients to various antigens has been demonstrated in several studies [26][27][28][29]. Moreover, the process of non-enzymatic glycosylation to the inciting antigens of GCA due to the hyperglycemic state might decrease their antigenicity and capability to activate dendritic cells [6].
Interestingly, a recent retrospective cohort study has demonstrated no protective benefit of DM against GCA [30]. In fact, the incidence of GCA was higher in the DM cohort. This study, however, was a coding-based study utilizing the billing records of Medicare claim data, unlike the primary studies included in this meta-analysis that identified their cohorts from temporal artery biopsy data and medical record review. Coding-based study is inherently at risk of coding inaccuracy and it is possible that the study might overestimate the incidence of GCA because of the lack of rigorous case verification.
A recent meta-analysis found no increased risk of coronary artery disease (CAD) among patients with GCA [7]. This result was somewhat unexpected as an increased CAD risk has been well documented in several chronic systemic inflammatory disorders, such as rheumatoid arthritis, systemic lupus erythematosus, systemic sclerosis, and inflammatory myositis [31][32][33][34]. This might be, in part, explained by the lower baseline prevalence DM, one of the strongest cardiovascular risk factors, as demonstrated in this study. This lower baseline risk might offset  the detrimental effect of chronic inflammation on endothelial cells of the coronary arteries. Nonetheless, we acknowledge the limitation that confounders might play a role in this association. We also could not evaluate the temporal relationship between GCA and DM. Thus, further prospective longitudinal studies are required to establish the role of DM as a protective factor against GCA.

Conclusion
In conclusion, our meta-analysis demonstrated a statistically significant lower baseline prevalence of DM among patients with GCA compared with sex-and age-matched controls. Whether DM has a protective role against the pathogenesis of GCA needs further investigation.