Author + information
- Received December 13, 2016
- Accepted December 14, 2016
- Published online March 6, 2017.
- Ängla Mantel, MDa,∗ (, )
- Marie Holmqvist, MD, PhDa,
- Daniel C. Andersson, MD, PhDb,c,d,
- Lars H. Lund, MD, PhDb,c and
- Johan Askling, MD, PhDa,e
- aDepartment of Medicine Solna, Clinical Epidemiology Unit, Karolinska Institutet, Stockholm, Sweden
- bDepartment of Medicine Solna, Unit of Cardiology, Karolinska Institutet, Stockholm, Sweden
- cDepartment of Cardiology, Karolinska University Hospital, Stockholm, Sweden
- dDepartment of Physiology and Pharmacology, Karolinska Institutet, Stockholm, Sweden
- eDepartment of Rheumatology, Karolinska University Hospital, Stockholm, Sweden
- ↵∗Address for correspondence:
Dr. Ängla Mantel, Karolinska Institutet, Clinical Epidemiology Unit, Department of Medicine Solna, T2, Karolinska Universitetssjukhuset, SE-17176 Stockholm, Sweden.
Background It is unknown whether the increased risk of heart failure (HF) in rheumatoid arthritis (RA) is independent of ischemic heart disease (IHD).
Objectives This study sought to investigate the relative risk of HF overall and by subtype (ischemic and nonischemic HF) in patients with RA and to assess the impact of RA disease factors.
Methods Two contemporary cohorts of RA subjects were identified from Swedish patient and rheumatology registries and matched 1:10 to general population comparator subjects. A first-ever HF diagnosis (classified as ischemic HF or nonischemic HF based on the presence of IHD) was assessed through registry linkages. Relative risks for a history of HF before RA onset were calculated through odds ratios. Relative risks of incident HF in RA were calculated as hazard ratios (HRs).
Results By the time of RA onset, a history of HF was not more common in RA. In the new-onset RA cohort, the overall HRs for subsequent HF (any type), ischemic HF, and nonischemic HF were between 1.22 and 1.27. The risk of nonischemic HF increased rapidly after RA onset, in contrast to the risk of ischemic HF. High disease activity was associated with all HF types but was most pronounced for nonischemic HF. In the cohort of patients with RA of any duration, the HRs were between 1.71 and 1.88 for the different HF subtypes.
Conclusions Patients with RA are at increased risk of HF that cannot be explained by their increased risk of IHD. The increased risk of nonischemic HF occurred early and was associated with RA severity.
The increased risk of cardiovascular disease in general and ischemic heart disease (IHD) in particular is well established in patients with rheumatoid arthritis (RA) (1). Considerably less is known about the association (if any) between RA and heart failure (HF). Retrospective cohort studies have suggested a doubled incidence of HF overall (2,3), an increased prevalence of HF overall (4), and an increased risk of mortality attributable to HF in patients with RA (5). Online Table 1 contains a list of studies to date. Evidence also suggests a possible association between rheumatoid factor (RF) positivity and high RA disease activity that accompanies the increased HF risk (2,6,7).
HF is a complex clinical syndrome with several potential risk factors and causes, of which hypertension and IHD are the most common; however, the increased prevalence of hypertension and IHD in RA might not fully explain the increased risk of HF in RA (2,8). Inflammation could contribute via coronary artery disease (9), as well as through additional pathways, including direct toxic or adverse remodeling effects of proinflammatory cytokines on cardiomyocytes, microvasculature, and extracellular matrix (10). Hence, RA-related inflammation might be an important driver for developing HF, leading to an increased risk of heart failure on top of, or even in the absence of, any increased risk attributable to IHD.
Previous studies of HF in RA have been based on RA cohorts of limited size and have exclusively assessed HF overall rather than by subtype or presumed origin. Therefore, this study aimed to assess the risk of incident HF overall and by subtype (specifically operationalized as HF occurring [or not] against a background of IHD) in large contemporary cohorts of patients with RA compared with the general population. We also sought to evaluate the relative risk (RR) of HF in time periods before and after RA onset and to investigate the impact of RA disease activity.
A unique personal identification number is assigned to all Swedish residents at birth or immigration and can be used to link information from official registries and other data sources (11). Several nationwide registries and 1 quality-of-care registry were used to detect exposures, outcomes, and covariates. The National Patient Registry (NPR) contains both inpatient and outpatient data since 1987 and 2001, respectively. Diagnoses are coded based on the Swedish version of the International Classification of Diseases (ICD) system (version ICD-10 in use since 1997). NPR was used to identify RA subjects (irrespective of disease duration) and baseline comorbidities of interest, as well as to detect the outcome of HF during follow-up. The Prescribed Drug Register was initiated in July 2005 and contains information on dispensed drugs from all Swedish pharmacies; it was used to collect information on selected drugs, used as proxies for comorbidities, before the index date for the cohort of established RA and their matched comparators. ICD and Anatomical Therapeutic Chemical classification codes used are provided in Online Table 2. The Swedish Rheumatology Quality Register (SRQ) is a national quality-of-care registry, initiated in 1995, that includes more than 10,000 patients with new-onset RA. SRQ contains information on rheumatologist-assigned RA diagnoses, disease duration, inflammatory activity, and overall disease activity collected at pre-specified time points. SRQ was used to identify the study subjects with new-onset RA and to collect information on disease characteristics. The Total Population Register contains demographic information, such as sex, age, civil status, and dates of migrations, on all Swedish residents and was used to identify the general population comparator subjects and for data on vital status and residency during follow-up. The Database for Labor and Education is operated by Statistics Sweden and was used to retrieve information on years of education (categorized as <9, 10 to 12, and >12 years).
We identified all individuals 18 years of age or older with at least 2 visits at inpatient or outpatient specialist clinics with a listed diagnosis of RA in the NPR, including at least 1 visit at an internal medicine or rheumatology clinic, between 2006 and 2012. The positive predictive value of this diagnostic algorithm has been reported to be approximately 90% (12). An index date (i.e., start of follow-up) was defined as the second visit that listed RA.
Using the SRQ, all individuals ≥18 years of age with newly diagnosed RA (defined as <12 months with RA symptoms) between 1997 and 2012 were identified and included in the new-onset RA cohort. The index date was defined as date of diagnosis.
Using the Total Population Register, each unique RA patient was matched with general population comparator subjects. Individuals with RA were matched 1:10 on birth year, sex, and area of residency by the time of their first RA diagnosis. Comparator subjects had to be alive and without RA at the index date of their corresponding RA subject.
To determine whether any increase in HF risk was already present at RA diagnosis, we treated the new-onset RA cohort and its comparator subjects as a case-control study. HF overall was defined as a first-ever main diagnosis of HF (inpatient care or specialized outpatient care), coded according to ICD-10, before RA onset. Relative risks were calculated as odds ratios (ORs) using logistics regression adjusted for age and sex.
In assessing HF after RA diagnosis, we used cohort designs, monitoring individuals for new-onset HF from the index date in each cohort. All individuals with an HF diagnosis before the start of follow-up in each cohort were excluded. HF overall was defined as a first-ever main diagnosis (inpatient care or specialized outpatient care) of HF during follow-up, based on ICD-10 coding. Follow-up started at the index date and ended at the point of incident HF (outcome); December 31, 2012; death; or migration out of Sweden. Compared with medical chart data, a primary diagnosis of HF in the NPR has previously been reported to have a positive predictive value of 95% (13).
Because HF was assessed based on ICD-10 codes, information on ejection fraction was not available for all study participants. Because our aim was to assess whether HF risk was increased in individuals with no known IHD at baseline or during follow-up, HF was subtyped based on information of antedating or concomitant IHD. Ischemic HF was thus defined as: 1) a first-ever main diagnosis of HF in individuals with a history of IHD (irrespective of whether it was before or after the start of follow-up, or whether it was a primary or secondary diagnosis); or 2) a main first-ever diagnosis code specifically indicating ischemic HF. Follow-up ended at the first of fulfillment of the outcome definition; December 31, 2012; death; first migration out of Sweden; or first HF without any antedating diagnosis of IHD.
Nonischemic HF was defined as a first-ever main diagnosis of HF during follow-up in individuals without a history of IHD (main or any contributory diagnosis) before the start of follow-up or during follow-up. Follow-up ended at the time of HF outcome; December 31, 2012; death; migration out of Sweden; incident IHD diagnosis; or incident ischemic HF.
Baseline data are presented as numbers, proportions, means, and medians as appropriate. As a measure of RR, hazard ratios (HRs) were calculated using Cox regression. Crude HRs were based on time since start of follow-up and adjusted for sex, area of residency, and age at the index date.
For the prevalent RA cohort (for which follow-up started in 2006 or later, such that drug dispensation data were available for all subjects prior to start of their follow-up), we further adjusted for educational level and yes/no baseline cardiovascular comorbidities/treatments (hypertension, IHD, diabetes type 1 or 2, chronic obstructive pulmonary disease, heart valve disease or surgery, atrial fibrillation, renal failure, alcohol-related conditions, and use of nitroglycerin, warfarin, acetylsalicylic acid, calcium antagonists, diuretic agents, renin-angiotensin-system blocking agents, beta-blockers, lipid-lowering agents, insulin, and oral antidiabetic agents). Stratum-specific HRs were presented for RF status at index date and sex. Assessment of the proportional hazard assumption was performed graphically with cumulative incidence functions by calculating stratum-specific HRs for different time periods since RA diagnosis and by introducing an interaction term between exposure (RA) and time of follow-up. The proportional hazard assumption was not violated in any of the analyses (p value between 0.09 and 0.40). To assess the impact of RA disease duration, subjects with new-onset RA and their comparators were stratified into periods based on length of follow-up, which was stratified as >1 year, 1 to 5 years, 5 to 10 years, and >10 years, with separate HRs calculated for each time period.
To refine our findings, we also assessed the impact of RA disease characteristics on the short-term risk of HF in RA cases with new-onset RA. Mean reported erythrocyte sedimentation rate, C-reactive protein (CRP), disease activity score, global health, and health assessment questionnaire between index date and end of follow-up were calculated and assessed with Cox regression models adjusted for age and sex. Similarly, we assessed oral corticosteroid use within the 2 months before HF or at the end of follow-up and biologic treatment, as a time-dependent variable, between index date and end of follow-up. For statistical power, HF was used whether listed as a main or contributory diagnosis (Online Table 3). All analyses were performed with SAS software package version 9.3 (SAS Institute, Cary, North Carolina). The study was approved by the ethics committee in Stockholm, Sweden.
We performed the following sensitivity analyses: 1) the RR of incident HF was assessed with not only main but also contributory diagnoses; 2) the RR of HF was assessed excluding all study subjects with any of the comorbidities or medications in Table 1 (i.e., in individuals without any registered hospitalization for or medication for cardiovascular disease); 3) nonischemic HF was redefined as no IHD allowed for up to 30 days after the HF event; and 4) ischemic HF was redefined so that IHD within 30 days after the HF was allowed to qualify as ischemic HF.
Characteristics of the study population are described in Figure 1 and Table 1. The median follow-up in all cohorts was approximately 5 years. Median disease duration (based on the time from the first recorded RA diagnosis until index date) was 3.4 years (interquartile range: 0.3 to 7.7 years) in the cohort with RA of any disease duration. RF-positive RA was detected in 72% of subjects in the cohort with RA of any duration and 66% of the new-onset RA subjects.
Hypertension, IHD, diabetes mellitus, and chronic pulmonary disease were more common in the RA cohorts, which also had a higher prevalence of prescriptions for calcium antagonists, diuretic agents, renin-angiotensin system–blocking drugs, β-blockers, and insulin (Table 1).
HF by cohort
By the time of RA diagnosis, a history of HF was not more common in subjects with new-onset RA than in their comparator subjects. Forty-one (2.8%) of the RA subjects and 217 (2.5%) of their comparators had a registered diagnosis of HF before the start of follow-up (OR: 1.01; 95% confidence interval [CI]: 0.92 to 1.14). Stratification did not reveal any difference; approximately 1% of RA subjects and their comparators had a history of nonischemic HF (OR: 0.98; 95% CI: 0.86 to 1.13), and 1.5% had a history of ischemic HF (OR: 1.01; 95% CI: 0.92 to 1.31).
In the new-onset RA cohort, 332 RA subjects (2.6%, or 4.1 per 1,000 person-years) versus 2,325 of their comparator subjects (2.0%, or 3.2 per 1,000 person-years) developed HF overall during follow-up (HR: 1.22; 95% CI: 1.09 to 1.37). In total, 152 (1.2%, or 2.6 per 1,000 person-years) versus 1,025 (0.9%, or 1.4 per 1,000 person-years) developed ischemic HF (HR: 1.27; 95% CI: 1.07 to 1.51), and 184 (1.5%) versus 1,312 (1.2%) developed nonischemic HF (HR: 1.22; 95% CI: 1.04 to 1.43). RF positivity was associated with a 40% higher risk of HF overall compared with RF-negative RA, for which there was no increase (RF-positive RA, HR: 1.45; 95% CI: 1.19 to 1.78 vs. RF-negative RA, HR: 0.97; 95% CI: 0.73 to 1.28; pinteraction = 0.02). Similarly, RF positivity was associated with a higher risk of ischemic HF than RF-negative RA (pinteraction = 0.02). For nonischemic RA, we noted no difference in risk by RF status. Stratification by sex did not reveal any differences in RR of HF overall or ischemic HF in particular, but there was increased risk of nonischemic HF among males compared with females (pinteraction = 0.03) (Table 2).
HF risk by RA duration
The RR of both HF types increased within the first year after RA diagnosis. In particular, the RR of nonischemic HF increased rapidly within the first year to an RR of ∼2 (HR: 2.06; 95% CI: 1.37 to 3.20) and thereafter declined over the different time periods. As for ischemic HF, risk remained and increased after 10 years with RA (Table 3). During the first year after RA onset, a higher mean erythrocyte sedimentation rate and disease activity score were associated with an increased risk of all HF types, but the association was more pronounced for nonischemic than ischemic HF. High CRP and high global health were associated with HF overall and nonischemic HF but not with ischemic HF. Oral corticosteroid use within the 3 months before HF was not associated with any increased risk of HF overall or ischemic HF but was associated with a tripled risk of nonischemic HF. Use of a biologic drug was not associated with an increased risk of any HF subtype (Figure 2).
In the cohort with RA of any duration, HF occurred in 1,129 patients (2.5%, or 6 per 1,000 person-years) versus 5,193 of the matched population comparators (1.4%, or 3 per 1,000 person-years; HR: 1.77; 95% CI: 1.69 to 1.85). A total of 529 RA subjects (1.2%, or 2.7 per 1,000 person-years) versus 2,883 comparator subjects (0.8% or 1.4 per 1,000 person-years) developed ischemic HF (HR: 1.88; 95% CI: 1.71 to 2.0). In total, 596 (1.4%, or 3.5 per 1,000 person-years) and 2,883 (0.8%, or 1.9 per 1,000 person-years) RA versus comparator subjects were registered with incident nonischemic HF (HR: 1.71; 95% CI: 1.57 to 1.87). Stratification by sex and RF status did not reveal any significant differences for any of the HF types. Adjustment for comorbidities did not appreciably alter the HRs (Table 4).
The exclusion of subjects with a registered diagnosis of any of the selected morbidities or dispensed drugs of any selected pharmacotherapy indicating cardiovascular disease (as reported in Table 1) left a population with no registered cardiovascular disease, which reduced the number of eligible subjects. If anything, this resulted in somewhat more pronounced HRs for all HF types during the complete follow-up period, as well as in the analysis of incident HF within 1 year from RA onset (Online Tables 4 and 5). Alternative definitions of incident HF did not appreciably alter the RRs (Online Tables 6 and 7).
Our population-based study of 2 large contemporary cohorts of subjects with RA and matched comparator cohorts is, to the best of our knowledge, the largest assessment to date of the risk of HF and the first study investigating HF risk by subtype. We observed an increased risk of HF overall, both in the presence and in the absence of IHD, which could not be readily explained by an increased occurrence of cardiovascular or other comorbidities in RA. Furthermore, our results showed no increased risk of HF before RA onset, but rather a rapid increase thereafter, with certain differences in development over time between subtypes (Central Illustration). Interestingly, we showed that this rapid increase in HF risk was associated with high RA disease activity.
Previous studies assessing the RR of HF in general in patients with RA have reported risk increases between 40% and 100% (2–4), comparable to our RR of 1.72 for HF overall. Contrary to most previous assessments, we further adjusted (and restricted) our RRs based on comorbidities and pharmacotherapies; we found that RRs remained elevated after such adjustments, and importantly, also when the population was restricted to individuals free of any of these comorbidities and medications.
This registry-based study was derived from HF-related data registered by physicians but without detailed information on actual clinical signs, laboratory measurements, or findings from examinations. Hence, we could not stratify HF into commonly used classifications such as whether ejection fraction was reduced or preserved. Instead, we focused on HF in the presence or absence of known IHD, which we could access from our data sources and which was also sufficient to investigate our primary aim to determine whether patients with RA seem to suffer an increased risk of HF regardless of their established increased risk of IHD. We could not, however, identify the specific cause of the increased HF risk among our RA subjects.
RA-related inflammatory activity could influence HF risk via several routes. Inflammation participates in all stages of the atherosclerotic process (14), and elevated inflammatory markers, such as CRP, have been associated with IHD both in the general population (15) and in patients with RA (16–19). Because RA is associated with an increased incidence and extent of IHD, including acute coronary events (1), finding an increased risk of ischemic HF at a level on par with that of the RR for IHD was perhaps not surprising. By contrast, the increased risk for HF in the absence of IHD was an important novel finding, raising the question whether the same or other RA-related factors drive the risk of this HF phenotype.
Previous studies have described a higher risk of HF for RF-positive versus RF-negative RA subjects (2,6). We extended this observation by making the same observation for ischemic HF, but without any such risk increase for nonischemic HF, which suggests that the increased risk of the former might be mediated via IHD, for which the risk increase is typically higher in seropositive RA (17,18).
Similar to what we have observed for acute coronary events in the same cohorts (20), we noted no increased risk before RA diagnosis but a rapid increase in risk thereafter (21). This was most pronounced early for nonischemic HF; excess risk was less pronounced thereafter. To investigate the role of acute effects temporally related to RA onset, such as uncontrolled inflammation, we assessed the association of disease factors in the risk of incident HF within the first year after RA onset. Our finding of an association between inflammatory activity and (in particular nonischemic) HF risk extended a previous indication of a link between RA disease activity and HF (6).
Even if we could not exclude differential diagnostic or coding intensity related to RA onset (vs. the general population), our results indicated an interesting link between high inflammatory activity and incident HF. Elevated inflammatory activity seen in conjunction with RA onset, before initiation of effective antirheumatic treatment, might affect the myocardium, leading to HF soon after RA diagnosis. In septicemia, cytokines such as tumor necrosis factor-α (TNF-α; involved in RA inflammation) are linked to reduced myocardial contractility within 10 min after exposure in vitro, as well as during longer time periods (22). Cardiomyocytes also react to inflammatory stimuli by expressing chemokines, cytokines, and cell-adhesion molecules, which leads to leucocyte recruitment and results in decreased cardiomyocyte contractility (23). Inflammation also induces endothelial dysfunction and hypertrophic and fibrotic responses in the myocardium, potentially leading to HF (24). Regarding the association between high disease activity and the risk of HF within the first year after RA onset, it is important to consider other potential risk factors, such as RA-related pharmacotherapies. As expected, we noted a strong association between use of oral corticosteroids (common in the initial management of RA in Sweden) and risk of nonischemic HF. TNF-α is elevated in humans with HF only, and the inflammatory response is thought to be a driver for HF disease progression. Therefore, the use of biological drugs, such as TNF-α antagonists, in RA has been a concern in the context of HF. Yet treatment with TNF-α antagonists in HF without overt autoimmune disease has shown no beneficial effect or even a worsening of symptoms in severe HF (25). Furthermore, a recent RA study of patients with and without prior HF found no increased risk of HF with TNF-α antagonists compared with traditional disease-modifying drugs (26). Moreover, TNF-α antagonists in RA can exert positive effects on cardiac function and disease biomarkers in HF (27,28).
In our study, RF-positive versus RF-negative RA was associated with a greater risk of ischemic HF. RF positivity is highly correlated with the occurrence of antibodies to anticitrullinated proteins (ACPAs) and is associated with greater RA disease activity and more inflammatory activity, at least if left untreated, but whether RF/ACPA per se is a risk factor remains unclear. In RA, RF is also linked to smoking (29). Citrullinated proteins are increased in the myocardial interstitium of patients with RA and have been associated with interstitial fibrosis (30). Moreover, a recent study on permeabilized cardiomyocytes found that citrullination can impair contractility by reducing myofilament Ca2+ sensitivity (31). Taken together, this suggests that inflammatory-mediated signals affecting the myocardium could potentially be involved in the pathogenesis of HF in patients with seropositive RA. In our study, we did not have access to data on ACPAs.
Our definition of nonischemic HF was based on the absence of a registered IHD diagnosis in the NPR. Importantly, however, the distribution of ischemic and nonischemic HF in our study was similar to that seen for ischemic versus nonischemic HF in general (32). Even though most IHD subtypes demand medical attention, and hence become registered, there are conditions, such as silent myocardial infarction and subclinical IHD, that might go unrecognized when using hospital data. Most IHD registrations consist of acute coronary events repeatedly proven to have a high validity against medical record data (33,34). Patients with RA have been reported to experience more frequent silent myocardial infarctions (35), as well as more extensive subclinical IHD, than non-RA subjects (36,37), which could have led us to overestimate the rate of HF without IHD among our RA subjects. There are reports, too, that HF may be overdiagnosed in this setting given the presence of conditions causing noncardiac dyspnea, such as pulmonary disease, that might be associated with RA and mistaken for HF. Potentially, patients with RA are also more frequently under health evaluation, which could lead to a general (but time-varying) increase in the observed rate of HF diagnoses. We attempted to reduce these potential biases by using only a main diagnosis of HF, known to have high validity (12), while avoiding a prevalent and concomitant HF that is diagnosed in conjunction with a visit for RA.
This study had several strengths. We accessed 2 large, contemporary cohorts of patients with RA, including 1 nationwide cohort consisting of unselected patients with prevalent RA of any duration. The positive predictive value for the exposure and outcome (here, RA and HF overall) has been proven high. The registries used to detect exposure, outcomes, and covariates of interest are based on prospectively collected data with high external validity and generalizability (38) in patients with almost complete coverage.
Patients with RA are at increased risk of both ischemic and nonischemic HF, which does not appear to be readily explained by established risk factors for HF. The risk increase, especially in the absence of IHD, develops rapidly after RA onset and is associated with high inflammatory and disease activity. From a clinical point of view, our findings emphasize the importance of clinical awareness and adequate assessment of HF in patients with RA.
COMPETENCY IN MEDICAL KNOWLEDGE: The risk of HF increases early after RA onset and is associated with the severity of inflammation.
TRANSLATIONAL OUTLOOK: Further studies should explore the clinical phenotype and prognosis of HF in patients with RA and the mechanisms by which inflammation and related factors contribute to cardiac decompensation.
For supplemental tables, please see the online version of this article.
Dr. Andersson was supported by the Stockholm County Council (clinical postdoctoral appointment) and the Swedish Heart Lung foundation. Dr. Lund was supported by the Swedish Research Council, the Swedish Heart Lung Foundation, and the Stockholm County Council; has received grants and speaker's honoraria from AstraZeneca and Novartis; and has served as a consultant to AstraZeneca, Novartis, Relypsa, and Vifor Pharma. Dr. Askling was supported by ALF, the Swedish Research Council, the Swedish Cancer Society, and the Swedish Foundation for Strategic Research; and has received research grants from Abbvie, Pfizer, Union Chimique Belge, Roche, Merck, Janssen, and AstraZeneca. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- anti-citrullinated protein
- confidence interval
- C-reactive protein
- heart failure
- hazard ratio
- International Classification of Diseases
- ischemic heart disease
- National Patient Registry
- odds ratio
- rheumatoid arthritis
- rheumatoid factor
- relative risk
- Swedish Rheumatology Quality Register
- tumor necrosis factor
- Received December 13, 2016.
- Accepted December 14, 2016.
- American College of Cardiology Foundation
- Avina-Zubieta J.A.,
- Thomas J.,
- Sadatsafavi M.,
- Lehman A.J.,
- Lacaille D.
- Wolfe F.,
- Freundlich B.,
- Straus W.L.
- Myasoedova E.,
- Crowson C.S.,
- Nicola P.J.,
- et al.
- Schau T.,
- Gottwald M.,
- Butter C.,
- Zaenker M.
- Libby P.
- Mantel A.,
- Holmqvist M.,
- Nyberg F.,
- et al.
- Gonzalez-Gay M.A.,
- Gonzalez-Juanatey C.,
- Lopez-Diaz M.J.,
- et al.
- Ajeganova S.,
- Andersson M.L.,
- Frostegard J.,
- Hafstrom I.
- Goodson N.J.,
- Symmons D.P.,
- Scott D.G.,
- Bunn D.,
- Lunt M.,
- Silman A.J.
- Marchant D.J.,
- Boyd J.H.,
- Lin D.C.,
- Granville D.J.,
- Garmaroudi F.S.,
- McManus B.M.
- Lim S.L.,
- Lam C.S.,
- Segers V.F.,
- Brutsaert D.L.,
- De Keulenaer G.W.
- Dibbs Z.I.,
- Diwan A.,
- Nemoto S.,
- et al.
- Solomon D.H.,
- Rassen J.A.,
- Kuriya B.,
- et al.
- Kotyla P.J.,
- Owczarek A.,
- Rakoczy J.,
- Lewicki M.,
- Kucharz E.J.,
- Emery P.
- Heslinga S.C.,
- Van Sijl A.M.,
- De Boer K.,
- Van Halm V.P.,
- Nurmohamed M.T.
- Stolt P.,
- Bengtsson C.,
- Nordmark B.,
- et al.
- Fert-Bober J.,
- Giles J.T.,
- Holewinski R.J.,
- et al.
- Zannad F.,
- Briancon S.,
- Juilliere Y.,
- et al.
- Karpouzas G.A.,
- Malpeso J.,
- Choi T.Y.,
- Li D.,
- Munoz S.,
- Budoff M.J.