Article Text

Concise Report
Smoking as a risk factor for the radiological severity of rheumatoid arthritis: a study on six cohorts
  1. D P C de Rooy1,
  2. J A B van Nies1,
  3. M C Kapetanovic2,3,
  4. H Kristjansdottir4,
  5. M L E Andersson5,6,
  6. K Forslind6,7,
  7. D M F M van der Heijde1,
  8. P K Gregersen8,
  9. E Lindqvist2,3,
  10. T W J Huizinga1,
  11. G Gröndal4,
  12. B Svensson6,
  13. A H M van der Helm-van Mil1
  1. 1Department of Rheumatology, Leiden University Medical Center, Leiden, The Netherlands
  2. 2Section of Rheumatology, Department of Clinical Sciences, Lund University, Lund, Sweden
  3. 3Department of Rheumatology, Skåne University Hospital, Lund, Sweden
  4. 4Department of Rheumatology and Center for Rheumatology Research, Landspítali–The National University Hospital of Iceland, Reykjavik, Iceland
  5. 5R and D Centre, Spenshult Hospital, Oskarström, Sweden
  6. 6Section of Rheumatology, Department of Clinical Sciences, Lund University, Sweden for the BARFOT study group, Lund, Sweden
  7. 7Section of Rheumatology, Department of Medicine, Helsingborg's lasarett, Helsingborg, Sweden
  8. 8Feinstein Institute for Medical Research and North Shore–Long Island Jewish Health System, Manhasset, New York, USA
  1. Correspondence to D P C de Rooy, Department of Rheumatology, Leiden University Medical Center, PO Box 9600, Leiden 2300 RC, The Netherlands; d.p.c.de_rooy{at}


Background Smoking is a risk factor for the development of anti -citrullinated protein antibodies (ACPA) positive rheumatoid arthritis (RA). Whether smoking predisposes to severe joint damage progression is not known, since deleterious, protective and neutral observations have been made.

Objective To determine the effect of smoking on joint damage progression.

Methods Smoking status was assessed in 3158 RA patients included in six cohorts (Leiden Early Arthritis Clinic (Leiden-EAC), BARFOT, Lund, Iceland, NDB and Wichita). In total 9412 radiographs were assessed. Multivariate normal regression and linear regression analyses were performed. Data were summarised in a random effects inverse variance meta-analysis.

Results When comparing radiological progression for RA patients that were never, past and current smokers, smoking was significantly associated with more severe joint damage in Leiden-EAC (p=0.042) and BARFOT (p=0.015) RA patients. No significant associations were found in the other cohorts, though a meta-analysis on the six cohorts showed significantly more severe joint damage progression in smokers (p=0.01). Since smoking predisposes to ACPA, analyses were repeated with ACPA as additional adjustment factor. Then the association was lost (meta-analysis p=0.29).

Conclusions This multi-cohort study indicated that the effect of smoking on joint damage is mediated via ACPA and that smoking is not an independent risk factor for radiological progression in RA.

  • Rheumatoid Arthritis
  • Smoking
  • Outcomes research
  • Ant-CCP
View Full Text

Statistics from


The severity of joint damage in rheumatoid arthritis (RA) is highly variable between patients. Genetic factors are estimated to explain half of this variance1; environmental factors likely play a role as well. No clear environmental risk factors for joint damage progression have been identified.

Smoking has been implicated as one of the most important environmental risk factors for the onset of RA, especially for the anti-citrullinated protein antibodies (ACPA) positive subgroup.2–5 It has been hypothesised that smoking contributes to the development of RA related autoantibodies.4 Whether smoking influences the severity of RA as well, is less clear. Smoking was associated with more severe radiographic progression in a Swedish study,6 but this relationship has not been established in other cohorts.7–9 Intriguingly, in a North American cohort, smoking was shown to protect against joint replacement surgery,10 and a Swiss study showed a trend towards less progression of radiographic joint damage in heavy smokers.9 The presence of ACPA is associated with more joint damage. It is unclear whether smoking as such affects progression of joint damage or whether smoking induces ACPA production and thereby affects joint damage progression.

This study therefore aimed to determine the association between smoking and joint damage in RA and whether this association is mediated through ACPA. In total 9412 radiographs of 3158 RA patients, from six cohorts, were studied and the results were summarised in meta-analyses.

Patients and methods


Cohort 1 consisted of 703 Dutch RA patients included between 1993 and 2006 in the Leiden Early Arthritis Clinic (Leiden-EAC), a population-based inception cohort that is described more extensively elsewhere.11 Hands and feet radiographs were taken at baseline and yearly over 7 years (total number 3656, mean follow-up 4.9 years) and chronologically scored by one reader unaware of clinical data using the Sharp–van der Heijde score (SHS).12 The within-reader intra-reader correlation coefficient (ICC) was 0.91.11 Treatment strategies differed for different inclusion periods, as described in de Rooy et al.11 Smoking status (present/past/current) was assessed by questionnaires.

Cohort 2 contained 839 RA patients included between 1992 and 1999 from the BARFOT study, a Swedish multicentre observational study of patients with early (disease duration ≤1 year) RA.13 Clinical, laboratory and radiological assessments were performed at inclusion and after 1, 2 and 5 years. Hands and feet radiographs (total number 2870, mean follow-up 4.3 years) were SHS scored by two readers. The between-reader ICCs at baseline and 2 years were 0.93 and 0.94, respectively. At inclusion, no patient had received prior treatment with disease modifying antirheumatic drugs (DMARDs) or glucocorticoids. During follow-up, 213 patients participated in a 2-year randomised study on low dose prednisolone as an addition to DMARD therapy. Smoking status (present/past/current) was assessed by the rheumatologist at inclusion.

Cohort 3 consisted of 339 RA patients that were recruited from a practice in Wichita, Kansas.14 Serial hands radiographs (total 1062) were made during 15 years of follow-up. Cohort 4 consisted of 885 RA patients included in the National Databank of Rheumatic Diseases (NDB).15 Hands radiographs were made at a single time point. The radiographs of cohorts 3 and 4 were SHS scored by one experienced reader (ICC=0.98). All patients in these two cohorts developed RA between 1963 and 1999, thus in eras when early treatment and use of biologicals were uncommon.16 Smoking status was assessed by questionnaires as a binary variable indicating ever or never smoking.

Cohort 5 consisted of 265 patients from Iceland, referred to Landspítali–The National University Hospital of Iceland or the private clinic of Reykjavik between 1970 and 2008. Radiographs were made at a single time point. Joint damage was determined using SHS by two trained readers. The ICCs between and within readers were all >0.95. Smoking status (present/past/current) was assessed by questionnaires.

Cohort 6 consisted of 127 early RA patients from Lund, Sweden that were prospectively followed during 5 years.17 ,18 Patients were recruited during 1985–1989. Radiographs of hands and feet were taken at baseline and annually for 5 years. Radiographs were scored chronologically according to the Larsen score by one of two readers (ICC between readers 0.94).19 Smoking status (present/past/current) was assessed by the rheumatologist.

All cohorts had been approved by the local medical ethical committee and all patients had given informed consent. RA was defined according to the 1987 ACR criteria except for the Lund cohort, where the 1958 criteria were used.

Statistical methods

Radiographic scores were log10 transformed to approximate a normal distribution. In the Leiden-EAC, BARFOT, Wichita and Lund cohorts, repeated radiographs were available. For these cohorts multivariate normal regression analyses were used with the log10 transformed radiographic score as response variable, as described previously.20 In the NDB and Icelandic patients, one radiograph per patient had been made. Here the estimated yearly progression rate was calculated (total SHS divided by disease years since diagnosis at time of radiograph) and linear regression analyses performed. The statistical methods used are extensively described in Knevel et al20 and in the online supplementary file. In all cohorts the radiological progression scores were compared between smoking groups, resulting in a relative difference in radiological progression. This effect estimate had no units and could be compared between datasets. All analyses were adjusted for age and gender. In the Leiden-EAC adjustments were made for the inclusion period (1993–1995, 1996–1999, 1999–2006) as a proxy for treatment strategy as described previously.11 Also in the Iceland data, adjustments were made for inclusion before or after 2000, to correct for different treatment regimes. In the BARFOT cohort, adjustments were made for participating in a corticosteroid study. No treatment adjustments were made in the cohorts where all patients were included in or before 1999 and treatment effects were not observed. Subsequently, all analyses were repeated with ACPA as additional adjustment factor in the regression analyses.

When possible the effect of smoking was first assessed by comparing three categories (present/past/never smokers); for the two North American datasets only binary variables were available. The results of the individual cohorts were summarised in an inverse variance meta-analysis testing for random effects. In these meta-analyses, for all cohorts, present and past smokers (combined as smokers) were compared with never smokers. Regression analyses were done using SPSS V.20.0, meta-analyses were performed using STATA. Two-sided p values <0.05 were considered significant.


Table 1 presents characteristics of the patients from the different cohorts. Overall 52–69% of the RA patients smoked or had smoked.

Table 1

Baseline characteristics of the study-population

In the Leiden-EAC, smoking was significantly associated with the progression of radiographic joint damage (p=0.042, figure 1). When smoking was categorised in two groups, smokers had a 1.02-fold (1.00–1.04) times higher progression rate per year (p=0.07).

Figure 1

Joint destruction over 7 years of disease in 703 Leiden Early Arthritis Clinic rheumatoid arthritis patients according to their smoking status (present, past or never smoker). Active smokers had a 1.01 (1.00–1.02) times higher radiological progression rate per year compared to former smokers, who had a 1.01 (1.00–1.02) times higher progression rate per year follow-up compared to never smokers. SHS, Sharp–van der Heijde score.

In the BARFOT patients, when comparing present, past and never smokers, smoking was associated with more severe joint damage, an effect that was constantly present over time (p=0.015). When ‘ever smokers’ were compared with ‘never smokers’, significance disappeared (p=0.12).

When the progression rates were compared between present, past and never smokers in the Icelandic and Lund cohorts, no significant results were obtained (p=0.75 and p=0.53, respectively). When smokers were compared with non-smokers, no significant results were also obtained (p=0.45 and p=0.38, respectively).

For the North American cohorts only binary data were available. In both cohorts smokers had no significant difference in joint damage progression compared to non-smokers (p=0.16 for the Wichita cohort, p=0.77 for NDB).

However importantly, in all mentioned cohorts, except for the Icelandic dataset, the directionality of the effect was similar, with smokers having more severe joint damage progression. Subsequently the results on the analyses comparing past and present smokers with never smokers were summarised in a meta-analysis, showing that smoking was significantly associated with more radiologic progression (p=0.01, figure 2A).

Figure 2

Meta-analysis on the effect of smoking (assessed as past and present smokers vs never smokers) on joint damage progression in six cohorts. Depicted are the results of the individual cohorts and of the meta-analysis. The meta-analysis is based on a random effects model. The I-square was 0.0%, indicating little heterogeneity between the cohorts. (A) Meta-analysis without adjustment for anti-citrullinated protein antibodies (ACPA) status; (B) the analyses on all cohorts were also adjusted for ACPA. The effect sizes indicate the increase in progression rate of smokers compared to non-smokers. Analyses were done on log10 transformed data; the effects estimated were back transformed to the normal scale and presented here. For example, for Leiden Early Arthritis Clinic patients depicted in the figure, smokers had a 1.02-fold (1.00–1.04) times higher progression rate per year (p=0.07) than never smokers; this indicates 2% more severe joint damage per year of follow-up. In all datasets the relative difference in radiological progression between smokers and non-smokers was assessed, yielding per dataset an effect estimate without units and that could be compared between datasets, independent on whether the radiographs were initially scored according to the Sharp–van der Heijde score or Larsen method. The cohorts with repeated radiological measurements yielded more precise estimations of the progression rate, resulting in smaller standard errors and a higher weight in the meta-analysis.

As smoking predisposes to ACPA formation and ACPA is associated with more severe joint destruction, we repeated all analyses with ACPA status as additional covariate. No significant results were obtained in any of the cohorts. The directionality of the effects was diverse. Also in the meta-analysis, smoking was no longer associated with joint damage progression (p=0.29, figure 2B), indicating that the observed effect of smoking was mediated through ACPA formation.


In this study we aimed to determine whether smoking is associated with the severity of the course of RA, reflected by the severity of joint damage. In a meta-analysis combining the data of six cohorts, it was observed that smokers had more severe joint damage. However since smoking predisposes to ACPA development, the analyses were repeated with adjustments for ACPA. Then the association was lost, indicating that the effect of smoking on joint destruction is mediated via the development of ACPA.

Advantages of this study are the large number of radiographs and patients, predominantly recruited in eras when early and aggressive treatment was less common. Hence the disease course of many of these patients may be more reflective of the natural disease course compared to many currently treated RA patients. Treatment differences occurred in part of the cohorts studied; adjustments were made where appropriate. Some cohorts had serial radiographs and others single radiographs per patient. The former results in more precise estimations of the radiological progression rate, which is reflected by smaller CIs of the effect estimates (see figure 2).

Some of the patients in the BARFOT, Iceland and Wichita cohorts were also assessed in earlier studies on smoking and joint damage.6 ,8 ,10 Given that some previous studies had contradictory results, an advantage of the present study is that we could combine data from these and other cohorts.

Our study has some limitations. Smoking was mainly assessed by questionnaires at disease onset. It is possible that patients may have failed to recall their former smoking habits. We had no information on the number of pack-years or on smoking habits during the disease course. Finally, we could not differentiate between past and present smokers in some cohorts, hence in the meta-analysis on all cohorts smokers were compared with non-smokers.

Studies on environmental risk factors for joint damage are relevant as such factors are potentially modifiable. Given that the effect of smoking was mediated via ACPA and that ACPA development occurs in the preclinical phase of RA, the current data may imply that preclinical environmental factors influence the long-term disease outcome.


View Abstract

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:


  • Handling editor Tore K Kvien

  • Contributors DPCdR, TWJH, DMFMvdH and AHMvdH-vM designed the study. DPCdR and JABvN performed the statistical analyses. MCK, HK, MLEA, KF, PKG, EL, TWJH, GG, BS and AHMvdH-vM were responsible for data collection. DPCdR and AHMvdH-vM wrote the first version of the manuscript. All authors read and approved the final version of the manuscript.

  • Funding This research was financially supported by BBMRI-Nl, a Research Infrastructure financed by the Dutch government (NWO 184.021.007).

  • Competing interests None.

  • Ethics approval Medical ethical committees of participating centres.

  • 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.