Article Text

Download PDFPDF

Extended report
Association between tobacco smoking and response to tumour necrosis factor α inhibitor treatment in psoriatic arthritis: results from the DANBIO registry
  1. Pil Højgaard1,2,
  2. Bente Glintborg1,3,4,
  3. Merete Lund Hetland3,4,5,
  4. Torben Højland Hansen6,
  5. Philip Rask Lage-Hansen7,
  6. Martin H Petersen8,
  7. Mette Holland-Fischer9,
  8. Christine Nilsson10,
  9. Anne Gitte Loft11,
  10. Bjarne Nesgaard Andersen12,
  11. Thomas Adelsten13,
  12. Jørgen Jensen14,
  13. Emina Omerovic3,
  14. Regitse Christensen1,
  15. Ulrik Tarp15,
  16. René Østgård16,
  17. Lene Dreyer1,4
  1. 1Department of Rheumatology, Gentofte Hospital, Copenhagen, Denmark
  2. 2Department of Rheumatology, Frederiksberg Hospital, Copenhagen, Denmark
  3. 3Copenhagen Center for Arthritis Research, Center for Rheumatology and Spine Diseases, Copenhagen University Hospital Glostrup, Denmark
  4. 4The Danish Rheumatologic Database (DANBIO), Copenhagen University Hospital Glostrup, Denmark
  5. 5Department of Clinical Medicine, Faculty of Health and Medical Sciences, University of Copenhagen, Copenhagen, Denmark
  6. 6Department of Rheumatology, Holbæk Hospital, Holbæk, Denmark
  7. 7Department of Rheumatology, Esbjerg Hospital, Esbjerg, Denmark
  8. 8Department of Rheumatology, Svendborg Hospital, Svendborg, Denmark
  9. 9Department of Rheumatology, Aalborg University Hospital, Aalborg,Denmark
  10. 10Department of Rheumatology, Odense University Hospital, Odense, Denmark
  11. 11Department of Rheumatology, Vejle Sygehus, Sygehus Lillebælt, Vejle, Denmark
  12. 12Department of Infectious Diseases and Rheumatology, Rigshospitalet, Copenhagen, Denmark
  13. 13Department of Rheumatology, Helsingør and Hillerød Hospital, Hillerød, Denmark
  14. 14Department of Rheumatology, Køge Hospital, Køge, Denmark
  15. 15Department of Rheumatology, Aarhus University Hospital, Aarhus, Denmark
  16. 16Department of Rheumatology, Silkeborg Hospital, Denmark
  1. Correspondence to Dr Bente Glintborg, Copenhagen Centre for Arthritis Research, Centre for Rheumatology and Spine Diseases, Copenhagen University Hospital Glostrup, Copenhagen, Denmark; glintborg{at}dadlnet.dk

Abstract

Objectives To investigate the association between tobacco smoking and disease activity, treatment adherence and treatment responses among patients with psoriatic arthritis (PsA) initiating the first tumour necrosis factor α inhibitor therapy (TNFi) in routine care.

Methods Observational cohort study based on the Danish nationwide DANBIO registry. Kaplan–Meier plots, logistic and Cox regression analyses by smoking status (current/previous/never smoker) were calculated for treatment adherence, ACR20/50/70-responses and EULAR-good-response. Additional stratified analyses were performed according to gender and TNFi-subtype (adalimumab/etanercept/infliximab).

Results Among 1388 PsA patients included in the study, 1148 (83%) had known smoking status (33% current, 41% never and 26% previous smokers). Median follow-up time was 1.22 years (IQR 0.44–2.96). At baseline, current smokers had lower Body Mass Index (27 kg/m2 (23–30)/28 kg/m2 (24–31)) (median (IQR)), shorter disease duration (3 years (1–8)/5 years (2–10)), lower swollen joint count (2 (0–5)/3 (1–6)), higher visual-analogue-scale (VAS) patient global (72 mm (54–87)/68 mm (50–80)), VAS fatigue (72 mm (51–86)/63 mm (40–77)) and Health Assessment Questionnaire (HAQ) score (1.1 (0.7 to 1.5)/1.0 (0.5 to 1.5)) than never smokers (all p<0.05). Current smokers had shorter treatment adherence than never smokers (1.56 years (0.97 to 2.15)/2.43 years (1.88 to 2.97), (median (95% CI)), log rank p=0.02) and poorer 6 months’ EULAR-good-response rates (23%/34%), ACR20 (24%/33%) and ACR50 response rates (17%/24%) (all p<0.05), most pronounced in men. In current smokers, the treatment adherence was poorer for infliximab (HR) 1.62, 95% CI 1.06 to 2.48) and etanercept (HR 1.74, 1.14 to 2.66) compared to never smokers, but not for adalimumab (HR 0.80, 0.52 to 1.23).

Conclusion In PsA, smokers had worse baseline patient-reported outcomes, shorter treatment adherence and poorer response to TNFi's compared to non-smokers. This was most pronounced in men and in patients treated with infliximab or etanercept.

  • TNF-alpha
  • Psoriatic Arthritis
  • Smoking
  • Treatment
  • Outcomes research

Statistics from Altmetric.com

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.

Introduction

Tumour necrosis factor α inhibitors (TNFi) are effective therapies in patients with psoriatic arthritis (PsA) who have insufficient response to synthetic disease-modifying antirheumatic drugs (sDMARD).1 ,2 However, only ∼60% of patients achieve ACR20 response.3–5 Thus, it is important to identify potential response modifiers to facilitate a rational and effective individualised treatment strategy.5 ,6 The impact of tobacco smoking is of particular interest since smoking is a potentially modifiable lifestyle factor.

Smoking is a possible risk factor for developing PsA, but results are conflicting.7 ,8 However, little is known about the impact of smoking on disease activity and TNFi treatment response in PsA.9–13 In axial spondyloarthritis, smoking increases disease activity and radiographic progression and reduces quality of life.14–16 In rheumatoid arthritis (RA) smokers have higher disease activity,17–19 and TNFi treatment is less efficacious.6 ,20–23

The nationwide DANBIO registry includes data on patients with rheumatologic diseases treated with TNFi in Denmark. We have previously described demographics and outcomes in patients with PsA treated with TNFi.3 ,24 The aim of the present study was to investigate differences between smokers and non-smokers regarding disease activity, treatment responses and adherence rates in patients with PsA initiating their first TNFi therapy in routine care. Furthermore, to study if the impact of smoking was influenced by gender and TNFi drug type.

Patients and methods

The DANBIO registry covers >90% of Danish adults treated in routine care with biologics due to rheumatic disease.25–27 According to Danish legislation, the registration and publication of data from clinical registries do not require patient consent or approval by ethics committees. Physicians are recommended to report data prospectively by an online system at least biannually and when medication is changed (http://www.danbio-online.dk).28

Baseline demographics include smoking habits, age, gender, Body Mass Index (BMI), disease duration, previous or current treatment with methotrexate (MTX) or other sDMARD. Functional status and peripheral disease activity are monitored by Health Assessment Questionnaire (HAQ),29 the 28-joint Disease Activity Score (DAS28),30 C-reactive protein (CRP) level (normal range ≤10 mg/L), and visual analogue scales (VAS) for scores of pain, patient's global and fatigue. It is not explicitly registered in DANBIO whether PsA patients have axial disease. Due to a limited number of registrations of axial disease activity at baseline and follow-up, these data were not included in the present study.

By 1 January 2012, 1536 patients with a diagnosis of PsA according to the treating rheumatologist had been registered and treated with a biological drug (bDMARD). We excluded patients treated with golimumab (n=19), bDMARDs other than TNFi (n=42), patients participating in clinical trials (n=54) or not followed in DANBIO since start of their first TNFi (n=33), leaving 1388 patients in the study.

Tobacco smoking status

In this study, patients were divided into three groups according to smoking status: current (≥1 cigarette/day), previous and never smokers. In previous smokers, the number of years since smoking cessation was recorded. Smokers, who had stopped smoking the same year as they started TNFi, were defined as previous smokers (n=24).

Queries were sent to the departments regarding patients with incomplete data on smoking status. Information was then obtained from hospital files or by asking the patients.

Treatment adherence

Treatment adherence was calculated as the number of years individual patients maintained treatment. Start date was the date of the first given dose, and stop date was the date of the first missed dose. Temporary treatment interruptions (eg, due to infections or surgery) of ≤3 months’ duration were allowed. All observations were censored at April 20th 2012. Among patients with no follow-up since December 2011, data were censored according to the last visit registered.

Reasons for drug discontinuation are registered in DANBIO in pre-specified categories: lack of effect, adverse events, disease remission, pregnancy, surgery, cancer, death, infections, loss to follow-up and other reasons. In the following, reasons for discontinuation are divided into three categories: ‘adverse events’ (including infection, death or cancer), ‘lack of effect’ and ‘other’ (including pregnancy, surgery, loss to follow-up, remission and other reasons for discontinuation).

Treatment response

Disease activity was evaluated at baseline and after 3 and 6 months’ therapy. The baseline visit was defined as a visit within the time window that ranged from 5 days before until 6 days after initiation of therapy. For the 3 months’ visit, the time window was 10–17 weeks, and for the 6 months’ visit 18–32 weeks after treatment start. If more than one registration occurred within a given time window for an individual patient, the registration closest to the given time-point was selected for analysis. If a patient had no registrations within a given time window, data were registered as missing for the given visit.

Clinical response was evaluated as achievement of ACR20/50/7031 or EULAR-good-response.32 Arbitrarily, we classified patients as ‘responders’ if they achieved clinical response (yes/no) at the 3-months’ and 6 months’ visits compared to baseline. In case of missing data at either the 3-months’ or 6 months’ visit, one registration of clinical response was sufficient to characterise the patient as responder. Patients who stopped treatment within the first 3 months of therapy were considered non-responders (n=272).

Statistics

Statistical analyses were performed by SPSS (V.20.0, SPSS, Chicago, Illinois, USA). Demographic and descriptive data are presented by medians/IQR. Groups were compared by non-parametric tests (χ2, Kruskal–Wallis and Mann–Whitney tests). In all tests, p values <0.05 were considered statistically significant. Calculations were based on observed data and no imputation of missing data was performed.

Kaplan–Meier plots and log rank tests were performed for analyses of treatment adherence for current, never and previous smokers. By univariate and multivariate Cox regression analyses we studied the impact of smoking on treatment adherence and associated HRs. The assumption of proportional hazards in the Cox regressions models was not fulfilled for previous smokers compared with current and never smokers and, therefore, only the latter two smoking categories were included (figure 1A). Univariate and multivariate logistic regression analyses and ORs were calculated to identify the impact of smoking (current/never) on clinical responses. Previous smokers were included in subanalyses. The following baseline factors were considered a priori confounders and included in all multivariate analyses: age (in quartiles), gender, disease duration (in tertiles), calendar year of starting TNFi (in tertiles). Age, disease duration and year of treatment start were transformed into categorical variables to allow for possible non-linear effects. Baseline swollen joint count (SJC), MTX use (yes/no) and TNFi type (adalimumab/etanercept/infliximab) were considered potential confounders, and were added one by one to the multivariate model, but were only included if they altered the OR/HR of smoking by >10%. MTX use and TNFi type did not alter the OR/HR of smoking by >10% in any of the analyses (data not shown). Baseline VAS scores, CRP, BMI, tender joint count, HAQ, and DAS28 were considered intermediate variables between tobacco smoking and outcomes, and were hence excluded from multivariate analyses.

Additional stratified Kaplan–Meier plots, and multivariate Cox and logistic regression analyses according to (1) gender and (2) TNFi type were performed.

In the subanalysis of time to discontinuation due to adverse events, discontinuations due to ineffectiveness were treated as censored observations and vice versa.

Results

A total of 1388 bDMARD-naive patients initiating treatment with adalimumab, etanercept or infliximab as the first TNFi were included (table 1). Among 1148 patients (83%) with known smoking status, 33% were current, 41% never and 26% previous smokers. Patients with missing smoking information had lower BMI, younger age, longer disease duration, higher CRP, higher SJC, lower VAS global and fatigue scores, compared to patients with available smoking information (table 1). Thirty-four percent of women and 31% of men with known smoking status were current smokers.

Table 1

Baseline demographics, disease activity and reasons for terminating TNFi treatment according to smoking status at the baseline visit

At baseline, current smokers had shorter disease duration, lower BMI, higher HAQ, higher VAS fatigue and VAS global compared to previous and never smokers (table 1). Previous smokers were older than current and never smokers. The reasons for stopping TNFi treatment were independent of smoking status (table 1). Male current smokers had higher HAQ (1 (0.6–1.4) vs 0.8 (0.4–1.3), (median (IQR)), p=0.03) and shorter disease duration (3 years (1–10) vs 6 years (2–13), p=0.03) than male never smokers, whereas VAS scores, DAS28 and BMI were similar (all p>0.05). Female current smokers had lower BMI (25 kg/m2 (23–29) vs 28 kg/m2 (24–32), p=0.001), higher VAS fatigue (77 mm (60–89) vs 61 mm (40–79), p=0.003) and shorter disease duration (2 years (1–6) vs 4 years (1–8), p=0.02) compared to female never smokers, whereas VAS global, DAS28 score and HAQ scores were similar (all p >0.05).

The median follow-up time for all included patients was 1.22 years (IQR 0.44–2.96) (current smokers 1.05 years (0.41–2.69), never smokers 1.37 years (0.43–2.90), previous smokers 1.12 years (0.44–2.94)). The total follow-up time was 2790 patient years. Current smokers had poorer treatment adherence than never smokers and median time on TNFi treatment was 1.56 years (0.97–2.15) in current smokers vs 2.43 years (1.88–2.97) in never smokers (median (95% CI)) (log rank p=0.02) (figure 1A). In men, current smokers had poorer treatment adherence than never smokers (figure 1B), whereas, the association between smoking and treatment adherence was less pronounced among women (figure 1C).

Figure 1

Kaplan–Meier drug adherence curves according to: (A) smoking status, all patients (log rank 7.7, p=0.02). (B) smoking status (current vs never), men (log rank 6.3, p=0.01). (C) smoking status (current vs never), women (log rank 1.1, p=0.3). (D) smoking stop year among previous smokers (log rank 1.0, p=0.053).

In univariate Cox regression analysis, current smoking was associated with poorer treatment adherence (HR 1.29, 95% CI (1.08 to 1.55) vs never smokers). In gender-stratified analyses, the same association was found for male current smokers (HR 1.41 (1.08 to 1.84) vs male never smokers) but not for female current smokers (HR 1.14 (0.89 to 1.46) vs female never smokers). In multivariate Cox regression analyses adjusted for gender, age, disease duration and start year of TNFi, we found no significant difference in treatment adherence between current smokers and never smokers, neither overall (HR 1.18 (0.97 to 1.44)) nor in males (HR 1.22 (0.97 to 1.67)) or females (HR 1.14 (0.87 to 1.5).

In Kaplan–Meier analyses stratified according to TNFi drug type, estimated median survival time was poorer among current versus. never smokers in patients treated with etanercept (1.0 year (0.66 to 1.39) vs 3.5 years (2.6 to 4.4), median (95% CI), log rank p=0.01), while smoking had no impact on adherence in patients treated with infliximab (1.2 years (0.69 to 1.61) vs 1.5 years (1.1 to 1.9), p=0.3) or adalimumab (2.8 years (1.9 to 3.7) vs 2.4 years (1.5 to 3.4), p=0.3). Similar results were found in univariate Cox regression analyses (data not shown). In multivariate Cox regression analysis, there was a statistically significant interaction between smoking and TNFi drug type (p=0.04). In stratified analyses according to TNFi drug type, current smoking was associated with poorer adherence to etanercept (HR 1.74 (1.14 to 2.66) vs never smokers) and infliximab (HR 1.62 (1.06 to 2.48)), but not adalimumab (HR 0.80 (0.52 to 1.23)) (table 2).

Table 2

Impact of smoking on treatment adherence stratified by TNFi drug type

Univariate Cox regression analyses stratified according to stop reason showed a comparable effect of current smoking on drug termination due to ‘adverse events’ and ‘lack of effect’ (HR 1.32 (0.98 to 1.78) and (HR 1.28 (0.98 to 1.67), respectively, both p=0.07 compared to never smokers).

Changes between baseline and 3 months’ and 6 months’ disease activity were calculated for VAS patient's global, VAS fatigue, VAS pain, CRP, tender and SJC according to smoking status. There was a non-significant tendency towards a greater decline in CRP after 6 months among never compared to current smokers (4.5 mg/L (0.3–17) vs 3 mg/L (0–11), p=0.08), whereas, data were insignificant at 3 months (p=0.6). No significant differences were found for delta VAS scores or delta joint counts among current versus never smokers at 3 months and 6 months (all p>0.05, data not shown).

Current smokers had lower EULAR-good-response and ACR20/50 response rates than had never smokers, whereas, ACR70 response rates were similar (figure 2). Twenty-three percent of current smokers achieved a EULAR-good-response after 6 months compared to 34% of never smokers (p=0.01). The rates for ACR20 and ACR50 response were 24%/33% (p=0.04) and 17%/24% (p=0.04), respectively. These differences were mainly present among men with EULAR-good-response: 24%/42% (p=0.002); ACR20: 25%/41% (p=0.01) and ACR50 response rates: 21%/32% (p=0.05) (figure 2). In univariate logistic regression analysis, current smokers had lower odds of achieving EULAR-good-response (OR=0.6 (95% CI 0.4 to 0.9) vs never smokers, p=0.01) and ACR20 (OR=0.7 (0.4 to –0.9), p=0.04), ACR50 (OR=0.6 (0.4 to 0.9) p=0.05) and ACR70 (OR=0.6 (0.4 to 1.16), p=0.14) responses. In multivariate analyses, the negative impact of smoking on treatment responses (ACR20/50/70 and EULAR-good-response) only reached statistical significance in gender-stratified analyses, where smoking was associated with a lower EULAR-good-response rate in men (OR=0.5 (0.3 to 0.9), p=0.03 current vs never smokers). In logistic regression subanalyses stratified by type of TNFi, smoking status did not affect response rates (overall and by gender, all p>0.05).

Figure 2

Treatment response rates after 6 months treatment according to smoking status overall and stratified according to gender. p Values are current vs never smokers (Mann–Whitney). (A) EULAR-good-response rates. (B) ACR20 response rates. (C) ACR50 response rates. (D) ACR70 response rates. Y-axis: percentage of patients achieving response.

In subanalyses, previous smokers were included as an additional group. In Kaplan–Meier analysis, previous smokers initially had drug adherence similar to current smokers. Beyond ∼6 months, the drug adherence for previous smokers was intermediate to those for never smokers and current smokers (figure 1A). When previous smokers were stratified according to number of years since smoking cessation, the drug adherence improved with more years since smoking cessation (figure 1D). In men, previous smokers tended to have higher EULAR-good-response and ACR20 response rates than current smokers, and lower rates compared to never smokers (univariate logistic regression analyses, data not shown) (figure 2).

Discussion

In this observational study of 1388 PsA patients initiating their first treatment with a TNFi, one-third of patients were current smokers. Current smokers had higher HAQ and patient VAS scores, but shorter disease duration than never smokers upon initiation of treatment. Current smokers had reduced response rates, most pronounced in male smokers where the EULAR-good-response and ACR20 response rates were nearly halved compared to male never smokers. For previous smokers, the effect of smoking on drug adherence diminished with time and was equivalent to never smokers ∼4 years after smoking cessation.

The impact of smoking on disease activity and functional status is well described in RA17–19 but scarcely investigated in PsA. A cross-sectional study among 283 patients with PsA reported that current smokers had higher HAQ.9 It has been suggested that smoking alters illness behaviour and causes a more severe perception of musculoskeletal pain.19 This may lead to higher DAS28 scores among current smokers compared to non-smokers with a similar inflammatory activity.22 Thus, perhaps the DAS28 score should be interpreted with more caution among smokers. This may be especially relevant in patients with PsA, in whom the validity of the DAS28 measure is debated.33 ,34 We found shorter disease duration among current compared to never smokers upon start of TNFi, which may indicate a more aggressive disease course among smokers.18 However, objective markers of disease activity (CRP and SJC) were independent of smoking status. One may hypothesise that a worse disease perception or a poorer general health condition among current smokers contributed to earlier TNFi treatment. Alternatively, sDMARD therapy may be less effective in smokers.19

The differences in baseline disease activity and demographics according to smoking status might explain, at least in part, why current smokers had poorer treatment response and treatment adherence in univariate but not multivariate analyses. However, in multivariate subanalyses, smoking was associated with poorer treatment response in men and poorer treatment adherence among patients treated with etanercept or infliximab. The non-randomised study design implies that these findings must be interpreted with caution due to the risk of residual confounding or uneven distribution of baseline demographics. Two previous studies have described the impact of smoking on TNFi treatment response in PsA. An observational study of 440 patients found current smoking to be associated with shorter 3-year TNFi drug survival.12 A single-centre study of 78 TNFi-treated PsA patients reported smokers to have poorer treatment response and lower drug retention rates after 6 months’ treatment in univariate analyses.10 Further studies are needed to confirm the relationship between smoking and treatment outcome in PsA. In RA, several studies have reported poorer TNFi adherence and treatment response among smokers.6 ,20–23 It has been suggested that smoking causes higher levels of TNF-α and other inflammatory markers,22 ,35 ,36 altered bioavailability of antirheumatic drugs, cutaneous vasoconstriction and slower absorption from subcutaneous injection, or increased basal metabolic rate.20 ,37 ,38 Few studies analysed the impact of smoking according to type of TNFi, and found that smoking mainly affected infliximab treatment.6 ,22 This might be explained by differences in drug metabolism or formation of antichimeric antibodies.22 ,39

We found that previous smokers who had stopped smoking more than ∼4 years ago had nearly same drug adherence rates as never smokers. This may illustrate a gradual normalisation of pathological processes and smoking-related behaviour, and is noticeable, as tobacco smoking is a potentially modifiable lifestyle factor. Studies in RA have found previous smoking to have no21 or an intermediate17 impact on the effect of TNFi compared to never smoking, though the influence of the time since smoking cessation was not investigated.

The strengths of this study are the high external validity for routine care due to inclusion of an unselected nationwide population of patients with PsA and the long follow-up time. Our study also has limitations. Smoking status was retrieved cross-sectionally although smoking status might alter later on.21 An obvious misclassification occurs when previous smokers resume smoking during follow-up. However, the exclusion of previous smokers from all main analyses made this bias less important. Furthermore, we had no valid data on the number of package years and thus the potential dose-response relationship between smoking and outcome could not be investigated. In Denmark, heavy smokers are more often men.40 ,41 One might assume that the stronger impact of smoking among male patients is associated with greater exposure to tobacco. Smoking may be linked to comorbid disease, depression, socioeconomic and lifestyle factors, which all potentially affect baseline disease activity and treatment outcome.17 ,22 Psoriatic manifestations in, for example, skin and nails may influence the decision on when to start TNFi treatment and the evaluation of treatment effect. However, these data are not included in DANBIO, and this might have affected our results, as smoking is suspected to increase the severity of skin psoriasis and to decrease the effect of TNFi on psoriatic skin lesions.42–44

In conclusion, we found current smoking to have a negative impact on treatment duration and clinical response in TNFi treatment of PsA, most pronounced in men and among patients treated with infliximab and etanercept. The effects seemed partially reversible, which stresses the importance of smoking cessation programmes for these patients. Clinicians should beware that current smokers potentially have higher HAQ and VAS scores compared with non-smokers, and this might affect the clinical evaluation of this patient group.

Acknowledgments

Thanks to all the departments of rheumatology in Denmark for reporting to the DANBIO registry.

References

Footnotes

  • Handling editor Tore K Kvien

  • Bente Glintborg and Pil Højgaard share first co-authorship.

  • Contributors All authors have contributed to acquisition of data, revised the article for important intellectual content and given approval of this version of the article to be published. PH, BG and LD have given substantial contributions to the study design, interpretation of data and formation of the article content. PH has been responsible for the collection and analyses of data.

  • Competing interests MH-F: UCB, MSD, Roche: Consulting fees, speaking fees, honoraria. AGL: Abbvie: Advisory board, Wyeth: Investigator.

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