Article Text

Download PDFPDF

Extended report
Mortality following new-onset Rheumatoid Arthritis: has modern Rheumatology had an impact?
  1. Marie Holmqvist1,
  2. Lotta Ljung2,
  3. Johan Askling1
  1. 1 Clinical Epidemiology Unit, Department of Medicine Solna, Karolinska Institute, Stockholm, Sweden
  2. 2 Department of Public Health and Clinical Medicine/Rheumatology, Umeå University, Umeå, Sweden
  1. Correspondence to Dr Marie Holmqvist, Department of Medicine, Solna Clinical Epidemiology Unit, Karolinska Institute, T2 SE-171 76 Stockholm, Sweden; marie.holmqvist{at}


Objective To investigate if, and when, patients diagnosed with rheumatoid arthritis (RA) in recent years are at increased risk of death.

Methods Using an extensive register linkage, we designed a population-based nationwide cohort study in Sweden. Patients with new-onset RA from the Swedish Rheumatology Quality Register, and individually matched comparators from the general population were followed with respect to death, as captured by the total population register.

Results 17 512 patients with new-onset RA between 1 January 1997 and 31 December 2014, and 78 847 matched general population comparator subjects were followed from RA diagnosis until death, emigration or 31 December 2015. There was a steady decrease in absolute mortality rates over calendar time, both in the RA cohort and in the general population. Although the relative risk of death in the RA cohort was not increased (HR=1.01, 95% CI 0.96 to 1.06), an excess mortality in the RA cohort was present 5 years after RA diagnosis (HR after 10 years since RA diagnosis=1.43 (95% CI 1.28 to 1.59)), across all calendar periods of RA diagnosis. Taking RA disease duration into account, there was no clear trend towards lower excess mortality for patients diagnosed more recently.

Conclusions Despite decreasing mortality rates, RA continues to be linked to an increased risk of death. Thus, despite advancements in RA management during recent years, increased efforts to prevent disease progression and comorbidity, from disease onset, are needed.

  • early rheumatoid arthritis
  • outcomes research
  • arthritis

Statistics from


Rheumatoid arthritis (RA) is a chronic inflammatory disease complicated by comorbidity, primarily cardiovascular disease.1 At least historically, RA has been linked to premature death.2 During the last two decades, more intense treatment strategies and new therapeutic options have been introduced in the management of new-onset RA,3 and the average level of RA disease control has improved.4 ‘Modern rheumatology’ might thus have impacted the elevated mortality in RA. Demonstrating a reduction, or elimination, of the excess mortality in RA on a population level would be the ultimate evidence that collectively, all of the components comprising ‘modern rheumatology’ have paid off.

With few exceptions, however, studies on RA mortality published so far have been based on patients diagnosed with RA and followed for mortality more than 10 years ago.5–8 A meta-analysis of studies on incident RA diagnosed from 1953 to 2007 failed to demonstrate any obvious decrease in the excess mortality, at least in relative terms,5 although a decreasing mortality rate was observed (but during the same time period, mortality rates have decreased also in the general population).5 9 10

To our knowledge, only one study has addressed the overall mortality in incident RA among patients diagnosed with RA during the most recent decade.11 This study, following 21 000 new-onset RA from 1999 to 2014 for mortality for a mean of around 3 years after RA diagnosis, indicated decreasing rates as well as relative risks (RR) compared with the general population over time (RR 1.56 in patients diagnosed from 1999 to 2006; 1.29 in patients diagnosed from 2007 to 2014).11 The short follow-up prompts the question of whether better management of RA has impacted longer term risks.

The aim of this study was therefore to assess overall mortality in a clinical inception cohort of patients with RA compared with the general population, with particular emphasis on the development of absolute risk and RR of death as a function of both disease duration and calendar period of RA diagnosis. Acknowledging that RA is a heterogeneous condition, we further set out to investigate mortality risks with different RA phenotypes.



We performed a nationwide population-based cohort study of patients with newly diagnosed RA (defined as RA diagnosis within 12 months of patient-reported symptom onset, to allow for assessments of mortality in relation to RA disease onset rather than RA diagnosis) with individually matched general population comparator subjects, based on prospectively recorded register data.


The publicly funded Swedish healthcare system enables access to all healthcare services, including specialised care for chronic diseases such as RA, for all residents. The incidence of RA in Sweden is estimated to be 40 per 100 000 from 2006 to 2008,12 and the prevalence on 31 December 2007 was 0.77%.13 Patients with RA are diagnosed and cared for by rheumatologists or internists, typically in hospital outpatient settings. Prescribed drugs are subsidised and, beyond an upper annual limit of Kr2200 (approximately US$260), provided free of charge. All residents are assigned a unique personal identity number that can be used for linkage of different data resources including several national health registers of high quality.14

Study population

The new-onset RA cohort

The Swedish Rheumatology Quality (SRQ) Register was initiated in 1995 and includes individuals aged 16 years or older fulfilling the 1987 American College of Rheumatology criteria for RA.15 The current coverage (on a national basis) is estimated to be above 80% of all (new-onset and established) RA.16 The register collects information on age, sex, rheumatoid factor (RF) status, date of first symptom of RA, date of inclusion into the register and the personal identification number. This register also contains information on drug treatment and disease activity, for example, number of swollen and tender joints, erythrocyte sedimentation rate, serum concentration of C-reactive protein, patient and physician’s assessment of global disease activity, and Disease Activity Score 28-joint counts (DAS28), from the inclusion visit and onwards. For this study, we identified all individuals diagnosed with RA between 1 January 1997 and 31 December 2014 who were included in the register within 12 months of first symptoms of RA (n=17 512). Validations against other data sources suggest that less than 0.8% of all new-onset RA in SRQ represent prevalent RA misclassified as new onset (Daniela di Giuseppe, SRQ, 2017, personal communication). For this study, the date of inclusion into the register (typically the date of RA diagnosis) was used as index date.

A general population comparator cohort

For each unique patient with RA, we randomly selected up to five individuals from the Swedish Population Register (that includes all Swedish residents), matched on sex, year of birth and residential area (n=78 847). Each subject was assigned the same index date as their corresponding patient with RA.

Data sources used for follow-up and mortality

Using the personal identification number, we linked the cohort of patients with RA and the matched general population comparator cohort to the Population Register for which data were available through 31 December 2015. The Population Register includes information on deaths, emigration and immigration for the entire Swedish population. Through these linkages, all deaths, causes of death and emigrations during follow-up were detected. The nationwide and near complete coverage of the National Patient Register ensured very low (<1%, but formally not assessable) losses to follow-up.

Follow-up and occurrence of death and its causes

The RA cohort and the comparison cohort were followed from index date until death, emigration or 31 December 2015 whichever came first.

Statistical analyses

Descriptive baseline data were summarised and presented as proportions, means or medians as appropriate. Mortality was assessed by dividing the number of deaths with the corresponding person-years of follow-up. The HR of death in RA compared with the general population was calculated using Cox regression models adjusted for residential area, sex, year of diagnosis and age at diagnosis. Analyses were stratified by RF status (yes/no), sex (men/women), age at index date (quartiles), calendar period of index date (1997–2001, 2002–2006, 2007–2011, 2012–2015), time since start of follow-up (<1 year, 1 to <5 years, 5–10 years, >10 years) and DAS28 (≤3.2 and >3.2) at RA diagnosis. All analyses were carried out with SAS software package V.9.4 (SAS Institute) and STATA IC V.11.2. This study was approved by the Stockholm Ethics Review Board.


Sixty-eight per cent of the 17 512 patients with RA and of the 78 847 general population comparator subjects were women. The mean age at index date was 58 years (table 1). When we split our study period into four groups as defined by calendar period of RA diagnosis, the gender distribution was stable but the distributions of age (increasingly higher) and RF status (increasingly more seronegative RA) varied over time (table 1).

Table 1

Characteristics at diagnosis among patients with new-onset RA overall and by calendar period of RA diagnosis, and in matched general population comparator subjects, expressed in n (%) if not stated otherwise

Clinical RA characteristics

During the study period, the duration of RA symptoms at index date (diagnosis) decreased somewhat from 1997–2001 to 2012–2015 (table 2). The disease activity at index date and at the return visit at 3–6 months also declined modestly. The use of any disease-modifying antirheumatic drug (DMARD) within the first year after RA diagnosis was high, and the proportion prescribed glucocorticoids, methotrexate or biological drugs during the first year after RA diagnosis increased.

Table 2

Clinical characteristics of patients with new-onset RA at inclusion in the Swedish Rheumatology Quality Register, including data on treatments initiated within 1 year of inclusion, by calendar period of RA diagnosis

Mortality during follow-up

During 123 360 person-years of follow-up (median/interquartile follow-up=6.2/7.5 years) in the RA cohort, 2386 individuals died. In the comparator cohort (median/interquartile follow-up=6.1/7.1 years), 9850 individuals died during 549 769 person-years of follow-up. In both the RA and the general population comparator cohort, the mortality rate was higher among men, increased with age and follow-up time, but decreased markedly over successive calendar periods of start of follow-up (table 3).

Table 3

Number of events (n), PYR and incidence rates (N/1000 PYR) of death in 17 512 patients with new-onset RA identified between 1997 and 2015 and in an individually matched general population comparator cohort (n=78 847). HR (RR) and 95% CI with the matching factors taken into account and adjusted for age and educational level. Overall and by sex, rheumatoid factor status, DAS28 at diagnosis, age at diagnosis and calendar period of diagnosis

RR of death by calendar period and time since RA diagnosis

Overall, RA was not associated with an increased mortality compared with the general population, HR 1.01 (95% CI 0.96 to 1.06) (figure 1table 4). However, we noted statistically significantly increased mortality risks in seropositive RA, in women with RA and in individuals 53–72 years at RA diagnosis (tables 3 and 4).

Figure 1

The relative risk of death among patients with rheumatoid arthritis (RA), overall and by calendar period of RA diagnosis compared with the general population and presented as HRs (95% CI). HRs are presented overall and by RA disease duration (<1 year, 1 to <5 years, 5 to <10 years, and ≥10 years since RA diagnosis).

Table 4

Mortality rates expressed as deaths per 1000 person-years with 95% CIs (within parenthesis) in patients with rheumatoid arthritis/general population

Within each calendar period of RA diagnosis, the overall RRs decreased during the later years, HR 1997–2001=1.09 (95% CI 1.01 to 1.18), HR 2002–2006=1.02 (95% CI 0.94 to 1.10), HR 2007–2011=0.95 (95% CI 0.86 to 1.05) and HR 2012–2015=0.77 (95% CI 0.63 to 0.95), an effect that was due to shorter follow-up in later calendar periods. When calendar years of RA diagnosis and RA disease duration were cross-tabulated, we noted similar HRs per follow-up interval for each calendar period of RA diagnosis (figure 1), with decreased risks of death the first year after RA diagnosis, in all calendar periods. Thereafter, absolute risks and HRs increased with disease duration but were not statistically significantly increased until 5 years after RA diagnosis (figure 1, table 4).

The same pattern of lack of distinct secular trend in RRs emerged when the same analyses were performed separately for seropositive RA (for which the relative mortality was typically increased throughout the study period) and seronegative RA (for which the RRs were typically not increased), and by DAS28 at diagnosis (table 5 and online supplement ary file 1).

Supplementary file 1

Table 5

HR and 95% CI for death comparing patients with RF-positive and RF-negative patients with RA compared with the general population, adjusted for sex, residential area, year of diagnosis and age. Overall and stratified by calendar period of RA diagnosis and time since RA diagnosis

The distribution of causes of death changed somewhat during the study period, both for patients with RA and the general population. We note a decrease in deaths attributed to diseases of the circulatory system in both groups, and reciprocal increases in malignant neoplasms (online supplementary table 2).

Supplementary file 2

The survival curves and their 95% CIs are presented in figure 2. The absolute 1-year mortality in RA was 0.8% (vs 1.4% in the general population), the 5-year mortality was 5% (vs 6%), the 10-year mortality was 10% (vs 10%), and at 15 years after RA diagnosis the mortality was 13% (vs 12%).

Figure 2

Kaplan-Meier survival curve comparing patients with new-onset rheumatoid arthritis (RA, red line) with the general population (blue line). Time since start of follow-up in years on x-axis. Numbers below the table denote individuals at risk at start of each follow-up period, and number of deaths at the end of each period in the RA and in the general population cohort, respectively.


In this large mortality study of an inception cohort of patients with RA, we made the following important observations: (1) There was a steady decrease in mortality rates over calendar time, both in the RA cohort and in the general population. (2) Across all calendar periods of RA diagnosis, an initial mortality ‘deficit’ turned into an excess mortality 5–10 years after RA diagnosis. (3) Taking RA disease duration into account, there was no clearly discernable trend towards lower excess mortality for patients with RA diagnosed more recently, although for obvious reasons not all calendar periods of RA diagnosis could (yet) be followed for 5 or 10 years. (4) The mortality pattern was distinctly different for seropositive and seronegative RA, and was influenced by baseline disease activity. Patients with seronegative RA, and patients presenting with low disease activity at diagnosis, had a lower and slower level of risk increase, if any. For all of the above reasons, including the confounding effects of the different time axes under study (age, calendar period and time since RA diagnosis), the overall RR in our cohort (HR=1.01) should not be interpreted as a null finding.

‘Modern rheumatology’ including features such as more therapeutic options, more effective use of conventional as well as biological DMARDs, earlier diagnosis, earlier treatment start and increasing vigilance for RA-related comorbidities holds the potential to reduce or eliminate the excess mortality in RA. Yet, studies evaluating the mortality among patients with incident RA from recent years have shown conflicting results.6–8 11 17 In the UK primary care data study by Zhang et al and in the Canadian administrative data study by Lacaillle et al there were indications of declining (relative) risks of death already during the first 5 years of RA disease, over calendar time, but longer term data were not available.7 11 Our study does not immediately corroborate these findings, mainly because in our study there was no excess mortality during the first 5 years, in any of the calendar periods under study. Instead, and despite clear secular trends in increasing use of DMARDs and decreasing disease activity during our study period (table 1), our study suggests a slower, yet steady, development of an excess mortality thereafter, that is, in follow-up periods not covered by the UK and the Canadian studies.7 11 In the Dutch study by Radovits et al no improvement in absolute or relative mortality was observed among patients diagnosed with RA from 1985 to 2007 and followed for mortality through 15 March 2008. Rather, there were indications of an increasing mortality gap with the excess risk of death being observed after 10 years of RA disease duration (which is more in keeping with the risk trajectory in our study).6 In the UK study based on the Norfolk Arthritis Register (patients with RA diagnosed from 1990 to 2004 and followed through 2011), the excess in the 7-year mortality remained unchanged over time.8

Our observation that the mortality rates in RA have declined in parallel to those in the general population suggests that at least, patients with RA have benefitted from general improvements in public health to the same extent as the general population, but also that factors specific to patients with RA remain at play and maintain the mortality gap. In this regard, our observation based on an inception cohort from 1996 followed through 2015 extends those from a recent meta-analysis that compared mortality in patients with new-onset RA with start of inclusion between 1955 and 1995. Since the observed mortality declines in the meta-analysis were equally pronounced in RA and in the general population, the excess mortality in RA remained unchanged in relative terms (an RR around 2).5 The absolute mortality presented in the meta-analysis, and in other studies,7 11 17 is numerically in line with ours and demonstrates a decline in the RA population, as well as in the comparators, except in the study by Humphreys et al where unchanged absolute risk and RR were presented in patients diagnosed during 1990–1994, 1995–1999 and 2000–20048 (rate 21/1000 person-years, and RR approximately 1.2 during a maximum of 7 years of follow-up).

The observation of an initial decrease in mortality following RA diagnosis may seem counterintuitive, has been observed in other studies of RA mortality (presented in abstract form only)18 and remains to be fully explained. Tentative explanations include diagnostic bias (lower tendency to diagnose RA in individual terminally ill from other diseases) and selection bias (lower likelihood of including patients into a longitudinal monitoring system if they are unlikely to survive until or attend at least the first 3 months’ return). Importantly, such initial selection is an unlikely explanation for the ensuing excess mortality as it would, if anything, reduce the observed mortality in RA, nor is it a likely explanation for the lack of improvement in excess mortality over calendar time as there is no sign of any time trend regarding this effect.

In addition to temporal trends for RA overall, we also observed quite different risk trajectories for seropositive and seronegative RA, with a slower and less pronounced development of excess mortality for the latter. An increased risk in seropositive RA was observed already after 5 years (or earlier) of disease, but an excess risk was noted also among the patients with seronegative RA, although later in the disease course. Indeed, seropositivity has been associated with a worse prognosis of the RA disease, with a higher disease activity over time.19 Of note is also our finding of no excess mortality, in any time window, among patients with low disease activity at diagnosis. These findings extend those from the studies by Humphreys et al and Radovits et al in which seropositivity and higher disease activity, respectively, were predictors of worse survival.6 8

Our study has several limitations. Inherent in all studies on time trends is that, for example, 5 and 10-year risks can only be calculated among individuals who were diagnosed 5 and 10, respectively, years ago. Therefore, we could not address other than shorter term risks for the most recently diagnosed patients. It is possible that the treatment paradigms established in recent years, and so clearly visible in the descriptive characteristics of our study cohort, will have a more demonstrable impact with longer disease durations. We set out to assess trends in mortality rather than to attribute any such to different causes. To perform any analyses of predictors of mortality therefore fell outside the aim of this study; we had limited access to data on lifestyle factors and did not include or accommodate data on comorbidities or other clinical phenotypes than seropositivity and disease activity at diagnosis.

Our study also has several strengths. Data on vital status for patients and referents could be linked from public mandatory registers with close to full coverage. The cohort was sufficiently large to enable analyses of subgroups as defined by age, gender, follow-up, calendar time and RA characteristics. Misclassification of prevalent RA as new-onset RA, which would introduce lead time bias in any ‘inception cohort’ study, is low in SRQ and thus unlikely to have influenced our results. The large, nationwide RA inception cohort including patients from small and large, public and private rheumatology units ensured a good generalisability.

In conclusion, despite a decreasing absolute mortality rate, the excess risk of mortality in RA compared with the general population remains, at least as far as this can currently be studied. To close this gap, increased efforts to prevent disease progression and comorbidity, from disease onset throughout the disease course, are needed.



  • Handling editor Tore K Kvien

  • Contributors Study concept and design: JA, MH, LL. Acquisition of data: MH, JA. Statistical analysis: MH, LL. Analysis and interpretation of data: MH, LL, JA. Drafting of manuscript: MH, LL. Critical revision of manuscript and final approval given: MH, LL, JA. Obtained funding: JA. Study supervision: JA.

  • Funding This work was supported by the Swedish Research Council, the Swedish Foundation for Strategic Research, Stockholm County Council (ALF), the Heart Lung Foundation, Karolinska Institutet (Strategic Research Area Epidemiology). Funders had no impact on the design or interpretation of the study or its results.

  • Competing interests LL has received personal fees for educational activities by Pfizer and Bristol Myers Squibb and has participated in advisory board arranged by Pfizer. JA has or has had research agreements with Abbvie, BMS, MSD, Pfizer, Roche, Astra-Zeneca, Lilly and UCB, mainly in the context of safety monitoring of biologics via ARTIS/Swedish Biologics Register. Karolinska Institutet has received remuneration for JA participating in ad boards arranged by Pfizer and Lilly.

  • Patient consent Not requested since this is a register based study

  • Ethics approval This study was approved by the Ethical Committee in Stockholm, Sweden.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.