Article Text

Download PDFPDF

Shared epitope defines distinct associations of cigarette smoking with levels of anticitrullinated protein antibody and rheumatoid factor
  1. Yuki Ishikawa1,2,
  2. Katsunori Ikari3,
  3. Motomu Hashimoto4,
  4. Koichiro Ohmura5,
  5. Masao Tanaka4,
  6. Hiromu Ito6,
  7. Atsuo Taniguchi7,
  8. Hisashi Yamanaka7,
  9. Tsuneyo Mimori5,
  10. Chikashi Terao2,8
  1. 1 Joslin Diabetes Center, Harvard Medical School, Boston, Massachusetts, USA
  2. 2 Laboratory for Statistical and Translational Genetics, Center for Integrative Medical Sciences, RIKEN, Yokohama, Japan
  3. 3 Department of Orthopaedic Surgery, Tokyo Women’s Medical University, Tokyo, Japan
  4. 4 Department of Advanced Medicine for Rheumatic Diseases, Kyoto University Graduate School of Medicine Faculty of Medicine, Kyoto, Japan
  5. 5 Department of Rheumatology and Clinical Immunology, Kyoto University Graduate School of Medicine Faculty of Medicine, Kyoto, Japan
  6. 6 Department of Orthopeadic Surgery, Kyoto University Graduate School of Medicine Faculty of Medicine, Kyoto, Japan
  7. 7 Institute of Rheumatology, Tokyo Women’s Medical University, Tokyo, Japan
  8. 8 Clinical Research Center, Shizuoka General Hospital, Shizuoka, Japan
  1. Correspondence to Dr Chikashi Terao, Laboratory for Statistical and Translational Genetics, Center for Integrative Medical Sciences, Riken, Yokohama 230-0045, Japan; a0001101{at}kuhp.kyoto-u.ac.jp

Abstract

Objects Although the association of cigarette smoking (CS) with susceptibility to rheumatoid arthritis (RA) has been established, the impact of CS on anticitrullinated cyclic peptide/protein antibody (ACPA) and rheumatoid factor (RF) levels in RA has yet been clear, especially in relation to shared epitope (SE) alleles.

Methods A total of 6239 subjects, the largest Asian study ever, from two independent Japanese cohorts were enrolled. Precise smoking histories, levels of ACPA and RF, and HLA-DRB1 allele status were withdrawn from databases. Associations between CS and high ACPA or RF levels, defined by the top quartiles, were evaluated. The effect of HLA-DRB1 alleles on the association was further investigated.

Results CS at RA onset conferred the risks of high levels of both antibodies, especially RF (OR 2.06, p=7.4×10–14; ACPA, OR 1.29, p=0.012), suggesting that RF level is more sensitive to CS than ACPA level. The patients who had quitted CS before RA onset showed a trend of decreased risks of developing high levels of ACPA or RF, and the risks steadily decreased according to the cessation years. The association of CS with high ACPA level was observed only in subjects carrying SE alleles, while the association of high RF level was observed regardless of SE.

Conclusions CS confers the risks of high autoantibody levels in RA in different manners; CS interacts with SE alleles on ACPA level, while CS impacts on RF level despite SE allele. These data suggest novel distinct production mechanisms of RF and ACPA.

  • rheumatoid arthritis
  • rheumatoid factor
  • anti-CCP
  • smoking
  • epidemiology

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.

Key messages

What is already known about this subject?

  • Cigarette smoking (CS) is a known risk of rheumatoid arthritis (RA) and development of anticitrullinated cyclic peptide/protein antibody (ACPA) and rheumatoid factor (RF).

  • Shared epitope (SE) alleles are strongly associated with seropositive RA and higher ACPA levels in European and Asian populations.

What does this study add?

  • CS affects not only positivity but also higher levels of both ACPA and RF.

  • The associations gradually decreased by smoking cessation depending cessation years.

  • The association of CS with high ACPA level is more apparent in the presence of SE alleles, while the risk of high RF levels is independent of SE presence.

How might this impact on clinical practice or future developments?

  • The importance of smoking cessation may be further highlighted especially in patients with SE alleles.

  • Research to address mechanisms of formation of RA autoantibodies should take into consideration difference between ACPA and RF especially in the context of SE.

Introduction

Rheumatoid arthritis (RA) is characterised by chronic inflammation and subsequent proliferation of synovial tissues leading to cartilage and bone destruction.1 Both environmental and genetic factors are known to contribute to the development of RA.2 3 Anticitrullinated cyclic peptide/protein antibody (ACPA) and rheumatoid factor (RF) are characteristic autoantibodies found in patients with RA. Previous studies have revealed that the positivity for and high levels of ACPA and RF are associated with joint destruction4–8 and systemic bone loss even in early phases of disease course,9 10 suggesting the importance of not only the positivity but also the levels of ACPA and RF on disease outcomes.

Cigarette smoking (CS) is one of the environmental risk factors for RA development.11 12 Males are reported to be more susceptible to CS on RA.12 CS is also known to affect both RF and ACPA formation.13 Importantly, both CS intensity and duration are directly related to the risk of RA development with prolonged increased risk even after CS cessation.12

HLA-DRB1 is the gene most strongly associated with RA.14 Most of the RA-associated HLA-DRB1 alleles share similar amino acid (AA) sequences at position 70–74 on HLA-DR β chain called the shared epitope (SE).15 Large-scale association studies have revealed that AA positions 11 or 13, 71 and 74 of HLA-DRB1 are strongly associated with RA in European population.16 17 A previous Asian study revealed a very similar genetic architecture to the European population, with AA position 57 unique to the Asian. HLA-DRB1 is also associated with positivity of RF and ACPA18 and the levels of ACPA,19 20 but not of RF.21 HLA-DRB1*09:01 and other alleles also showed associations of the levels of ACPA and the associations of HLA-DRB1 were mainly explained by 74th AA alanine.19 20

The association between CS and ACPA formation in the context of HLA-DRB1 alleles, especially SE alleles, has been investigated in several studies.22–27 Most of them including two Asian cohort studies reported an interaction between CS and SE with regard to ACPA-positive RA development with exception of single north American cohort. However, the interactive association of SE and CS with ACPA levels (not positivity) have not been well-studied. The impact of CS cessation on ACPA levels, instead of risk of RA, have never been explored as well. Furthermore, since most of the recent studies focused on ACPA, recent knowledge about associations of CS and SE with RF levels has been lacking.

In the present study, we investigated impacts of CS and its cessation, especially at the time of RA onset, on future levels of ACPA and RF in 6239 RA patients from two independent single centre cohorts, which was the largest Asian study ever. We also investigated the associations of HLA-DRB1 alleles, especially SE alleles and HLA-DRB1*09:01 allele, and AA position in HLA-DRB1 with high levels of autoantibodies in relation to CS.

Methods

Patients

The study participants were recruited from two independent prospective comprehensive single-centre cohorts (545 from Kyoto University Rheumatoid Arthritis Management Alliance: KURAMA cohort28 and 5694 from Tokyo Women’s Medical University Institute of Rheumatology, Rheumatoid Arthritis: IORRA cohort).29 All patients are Japanese RA patients, and each had a diagnosis of RA based on the 1987 American College of Rheumatology (ACR) criteria and/or 2010 ACR/European League Against Rheumatism classification criteria for RA.30 31

Detailed smoking history were collected from the participants using questionnaire sheets. Pack-years, a standardised numerical value of lifetime tobacco exposure, was calculated by multiplying the number of cigarette packs smoked per day by the number of years the individual has smoked (cigarette packs per day × years of smoking).

Written informed consent was obtained from each participant. This study was approved by local ethical committee of each institute (online supplementary notes). Clinical information was obtained from the database of the two cohorts in an unbiased manner.

Study design

A schematic view of the study designs are illustrated in online supplementary figure S1. In brief, patients were first stratified based on their smoking histories at the time of last visits. Patients who had ever smoked and who had never smoked were defined as ever-smokers and never-smokers, respectively. Then, ever-smokers were further stratified into three categories; smokers who had smoked at onset of disease (smokers at onset (SaO)), ex-smokers who quitted smoking before onset (ex-smokers at onset (exSaO)) and subjects who started smoking after onset (others).

Quantification of RF and ACPA

ACPA was quantified as second-generation anticyclic citrullinated peptide antibody by MesaCup CCP ELISA kit (Medical and Biological Laboratories).19 32 IgM-RF was quantified by latex-turbidimetric immunoassay, Iatro-RF II (Mitsubishi Kagaku Medicine).5 The cut-off levels of the antibodies were according to manufacturers’ instructions (ACPA <4.5 AU/mL; RF <20 IU/mL for IORRA, 15<IU/mL for KURAMA). High levels of ACPA (>236 AU/mL) or RF (>139 IU/mL) were determined according to the top quadrant levels within patients who were positive for these autoantibodies.

HLA genotyping

A WAKFlow system (Wakunaga) or an AlleleSEQR HLA-DRB1 typing kit (Abbott) was used for HLA-DRB1 typing, as previously described.32 HLA-DRB1 *01:01, *04:01, *04:04, *04:05, *04:10, *10:01, *14:02 and *14:06 were defined as the SE alleles.

Statistical analysis

Descriptive summary statistics are provided in table 1 for all continuous variables with parametric or non-parametric data as appropriate.

Table 1

Demographic features of subjects in each cohort

We selected the following items as candidates of basic covariates in the association studies for positivity or high levels of ACPA and RF, namely, age, age at onset of RA, disease duration, female gender and institutional variable. Disease duration was natural log-transformed due to their skewed distribution. Using these basic covariates, we first investigated the risks of positivity or high autoantibody levels in ever-smokers, SaO or exSaO using multiple logistic regression models. For high autoantibody levels, patients positive with ACPA or RF were included in the models. We also applied linear regression analysis rather than defining high/not-high autoantibody levels. Then, gender difference of the risk in each model was evaluated. Association of pack-years and their gender difference were analysed and expressed by β coefficients. Finally, we added presence of SE, HLA-DRB1*04:05, HLA-DRB1 non-*04:05 SE and HLA-DRB1*09:01 alleles in each model to adjust with these known risk-related HLA-DRB1 allele covariates. HLA-DRB1*09:01 allele is frequently observed in Japanese and strongly associated with lowering ACPA levels.33 Never-smokers were set as reference subjects in all the analyses. Omnibus test was performed as described previously.16 21 34

The effect difference of covariates from two separate regressions was calculated in the following manner; the results of two regressions were combined but the estimation was kept separate as a seemingly unrelated estimation (SUE), and the generalised Hausman test was conducted to obtain a one-sided p value (pSUE).

Binominal probability test was conducted to compare the effect sizes between males and females assuming chance of difference was 0.5.

Pearson’s correlation coefficient was calculated to test on linear trend of proportions.

Stringent significance levels were set based on Bonferroni correction. P values <2.4×10–4 (0.05/208) were regarded as significant in omnibus tests. For the other analyses, p values <0.05 were set as significant. The missing data (see also online supplementary data) were handled as missing completely at random due to the complete lack of relationship between missingness of the data and observed outcome variables (χ2 p>0.05). All the statistical analyses were performed on STATA/IC V.14.

Results

Enrolment of study participants and definition of subgroups

The demographic features and smoking information of enrolled subjects were summarised in table 1. We stratified subjects based on their smoking in the questionnaires as described in the Methods section.

Significant association between CS and positivity and high levels of ACPA and RF

Consistent with the previous studies, we found that SaO (ones who smoked at onset, see the Methods section) had higher risks of positivity of ACPA and RF (OR 1.39 (95% CI 1.09 to 1.76) and 1.52 (1.26 to 1.85), p=0.0068 and 1.8×10–5, respectively, online supplementary figure S2A). We also found a dose-dependency of the associations (p=0.0074 for both ACPA and RF; online supplementary figure S2B). When we focused on subjects positive for the autoantibodies, we found that SaO had significantly positive associations with high levels of ACPA and RF (OR 1.29 (1.06 to 1.57) and OR 2.06 (1.70 to 2.48), p=0.012 and 7.4×10–14, respectively, figure 1A) where the effect size of RF was higher than that of ACPA (pSUE=5.1×10–4 ; we also got comparable p-values by permutating obtained β coefficients based on their means and standard errors and comparing the difference of their distribution). Again, we observed dose-dependent association of CS with high RF levels (p=0.0065; figure 1B). These results suggest that CS has a larger effect on RF formation than on ACPA. Furthermore, most of the effect sizes were larger and significant in male than in female patients (binominal probability test: p=2.8×10–4), which was comparable to the previous studies of RA onset reporting males being more susceptible to CS.12

Figure 1

Cigarette smoking at the time of onset is a significant risk of high levels of ACPA and RF. The associations of smokers at the time of onset (SaO) with high levels of ACPA or RF were evaluated referring never-smokers (A). β coefficients of pack-years at the time of onset for high levels of ACPA or RF referring to never-smokers are presented (B). ORs in (A) and β coefficients in (B) are indicated by dots, and 95% CIs are indicated by two-sided lines. *p<0.05, **p<0.01, ****p<1.0×10–4. The numbers of patients, ORs and β coefficients were described in online supplementary notes. High ACPA or RF: top quartile of ACPA or RF-positive patients. ACPA, anticitrullinated cyclic peptide/protein antibody; RF, rheumatoid factor; SaO, smokers at onset.

We conducted the same analyses in ever-smokers, who were more heterogenous population than SaO and found that ever-smokers had almost the same risks as SaO (online supplementary figure S3A–D).

Smoking cessation before onset may lower risks of future high levels of ACPA and RF depending on the duration

Next, we assessed the association of CS cessation before onset with future ACPA and RF levels to evaluate attenuated influence of CS after cessation on autoantibody formation. We found that exSaO (ones who had quitted smoking before onset of RA, see the Methods section) had reduced risks of future positivity (ACPA, OR 1.29 (1.01 to 1.66), p=0.042; RF, OR 1.07 (0.88 to 1.30), p=0.50; online supplementary figure S4A) and of high levels of ACPA and RF (ACPA, OR 1.11 (0.88 to 1.41), p=0.39; RF, OR 1.23 (0.95 to 1.59), p=0.11; online supplementary figure S4B) compared with SaO or ever-smokers (online supplementary table S2). The effect sizes of smoking amount before smoking cessation were smaller than those of ever-smokers or SaO in the context of RF levels (online supplementary table S3, figure S4C, S4D), again indicating RF is more sensitive to CS than ACPA.

Stratifying the exSaO by their cessation years showed that the associations of CS with both positivity (figure 2A) and high levels of ACPA and RF (figure 2B) decreased depending on the cessation years. Likewise, percentages of patients with positive (ACPA, p=0.023; RF, p=0.037; figure 2C) and high levels of ACPA and RF (ACPA, p=0.52; RF, p=0.26; figure 2D) steadily decreased in the same manner.

Figure 2

Ex-smokers at onset had no longer significant risk of future high levels or positivity of ACPA and RF depending on the cessation years. Patients who had quitted CS before onset were stratified according to their cessation years, and the association of each category of patients with positivity (A) or high levels (B) of ACPA and RF were evaluated referring that of never-smokers. ORs are indicated by dots, and 95% CIs are indicated by two-sided lines. The numbers of patients and ORs were described in online supplementary notes. Percentages of the patients stratified by the cessation years are presented based on positivity (C) or levels (D) of ACPA and RF. **p<0.01. High ACPA or RF: top quartile of ACPA or RF positive patients. ACPA, anticitrullinated cyclic peptide/protein antibody; CS, cigarette smoking; RF, rheumatoid factor.

Together, these results indicated that exSaO had less risks of future positivity and high levels of ACPA and RF than ever-smokers or SaO, and the risks were gradually attenuated depending on duration of smoking cessation.

CS affects ACPA levels only in the presence of SE alleles

Finally, we evaluated the interactive association of CS and SE with ACPA and RF levels (online supplementary table S1). When conditioned on SE presence, the association between SaO and high ACPA level was no longer significant (preconditioned p=0.012, figure 1A; postconditioned p=0.38, figure 3A). To confirm the finding, we stratified the patients into four categories according to their smoking status and the presence of SE. Neither SE nor smoking history affected the proportions of patients (χ2 p=0.120, online supplementary table S4). Setting SE(-) never-smokers as a reference, both SE(+) SaO (OR 3.10 (1.91 to 5.03), p=4.7×10–6) and SE(+) never-smokers (OR 2.23 (1.59 to 3.14), p=4.3×10–6) had an increased risk of high ACPA level (figure 3B). Importantly, SE(-) SaO did not show even a trend of positive association, strongly indicating a SE-dependent association between smoking and high ACPA levels. On the contrary, the association between SaO and high RF level remained significant when conditioned on SE allele presence (p=0.0044, figure 3A). SE(+) never-smokers showed a rather negative association (figure 3B, indicated by green), while both SE(-) and SE(+) SaO showed a trend of positive associations (figure 3B, indicated by red and yellow, respectively), indicating an association of CS with high RF level is independent on SE. These association patterns were also observed in ever-smokers and in linear regression analysis where ACPA/RF levels were used as dependent variables (online supplementary figure S5).

Figure 3

Cigarette smoking affects ACPA levels only in patients with shared epitope alleles while cigarette smoking per se affects high RF levels regardless of shared epitope allele status. (A) The association of SaO with high ACPA or RF levels were evaluated conditioning on presence of shared epitope (SE) alleles. ORs are indicated by dots, and 95% CIs are indicated by two-sided lines. (B) The associations of SaO with high ACPA or RF levels with or without SE alleles were evaluated referring never-smokers without SE alleles. ORs are indicated by dots, and 95% CIs are indicated by two-sided lines. **p<0.01, ***p<0.001, ****p<1.0×10–4. The numbers of patients and ORs were described in online supplementary notes. High ACPA or RF: top quartile of ACPA or RF positive patients. ACPA, anticitrullinated cyclic peptide/protein antibody; RF, rheumatoid factor; SaO, smokers at onset.

Since we previously reported that among SE alleles HLA-DRB1*04:05 has unique features in the context of joint destruction20 ,35 and others reported strong interactive associations between CS and SE alleles (mainly non-*04:05 SE) on ACPA positivity in Europeans,23 27 we assessed whether the associations of SE with autoantibody levels in relation to CS were driven by HLA-DRB1*04:05. However, we did not find a specific association of HLA-DRB1*04:05 (or non-*04:05 SE) (figure 4A,B).

Figure 4

The comparative associations of cigarette smoking with HLA-DRB1*04:05 or non-*04:05 SE alleles on high levels of ACPA and RF. The associations between f high levels of ACPA or RF and SaO with or without HLA-DRB1*04:05 (A) or non-*04:05 SE allele (B) were evaluated referring never-smokers without HLA-DRB1*04:05 or non-*04:05 SE alleles. ORs are indicated by dots, and 95% CIs are indicated by two-sided lines. ***p<0.001, ****p<1.0×10–4. The numbers of patients and ORs were described in online supplementary notes. High ACPA or RF: top quartile of ACPA or RF positive patients. ACPA, anticitrullinated cyclic peptide/protein antibody; RF, rheumatoid factor; SaO, smokers at onset.

We further conducted omnibus tests to investigate associations between AA positions in HLA-DRB1 and high ACPA levels with (figure 5) or without (online supplementary figure S6A,B) conditioning on SaO. We found that AA position 74, a part of SE sequence, had the strongest association with high ACPA levels (pomnibus=4.5×10–12 with conditioning on SaO; pomnibus=1.1×10–13 without conditioning SaO; online supplementary table S5), which is comparable to our previous finding that AA position 74 is strongly associated with both high ACPA levels in ACPA-positive RA and RA susceptibility.20 While other AAs also showed significant associations (online supplementary table S5), these associations could be explained by AA position 74 (figure 5). On the other hand, none of the AA position in HLA-DRB1 was associated with high RF levels (online supplementary table S6C,D), also consistent with our previous study.21

Figure 5

Amino acid position 74 in HLA-DRB1 is most significantly associated with high ACPA levels regardless of smoking status of subjects. Omnibus p values are shown for each HLA-DRB1 amino acid (AA) position of subjects with anticitrullinated cyclic peptide/protein antibody (ACPA). The horizontal red lines indicate the level of significance (p=2.4×10–4). High ACPA or RF: top quartile of ACPA or RF positive patients. RF, rheumatoid factor; SaO, smokers at onset.

Based on the strongest and unique (a reduced effect size) association of HLA-DRB1*09:01 with levels of ACPA,19 20 we also evaluated associations among CS, HLA-DRB1*09:01 allele, and high autoantibody levels. However, we found no apparent associations driven by HLA-DRB1*09:01 (online supplementary figure S7).

Taken together, these results show that CS affects high ACPA levels only in the subjects with SE alleles, while CS per se affects high RF levels regardless of SE allele presence.

Discussion

In the present study, we investigated the impact of CS on ACPA and RF levels in the largest Asian study ever. We found that CS, especially at the time of onset, was a significant risk of high levels of both ACPA and RF in a dose-dependent manner. The effect sizes were larger for RF than ACPA suggesting that RF is more sensitive to CS, and were larger in male patients compared with female patients. These risks were attenuated by CS cessation depending on cessation years. The association of CS with ACPA levels was apparent only in the subjects with SE alleles, while CS independently affects RF levels regardless of SE, implying that interaction between CS and SE alleles can impact only on ACPA formation.

Considering that CS is one of risk factors of periodontal disease,36 our findings together with previous studies further support the idea that CS is strongly associated with RA-related autoantibody formation beyond ethnicities.13 Of note, we also showed that RF is more sensitive to CS than ACPA, suggesting that the mechanisms of ACPA and RF formation are different as indicated in the previous study.37

We showed that associations of CS with autoantibody levels were attenuated by CS cessation depending on the cessation years, but the association with ACPA levels still remained increased 20 years after CS cessation. These long effects of CS seem compatible with the previous study demonstrating that both smoking intensity and duration were directly related to a risk of RA development especially in RF-positive cases, with prolonged (~20 years) increased risk after cessation.38 The present study stood on intracase analyses lacking ACPA-positive or RF-positive non-RA subjects and thus could not draw any conclusions with respect to the association of CS with RA onset, but still might support the previous observation mentioned above with additional information of the impacts of CS on autoantibody levels.

SE and HLA-DRB1*09:01 are well-established genetic risks for ACPA-positive RA,34 39 and the latter is also known to lower ACPA levels in Japanese RA patients.33 Several studies have investigated associations of CS and SE allele presence with ACPA formation, and most of them confirmed the interactive effect of CS and SE on RA development especially in ACPA-positive individuals.22–27 In addition to these results, we found that CS affected high ACPA level only in the subjects with SE alleles, while the association of CS with RF levels was independent on SE. Furthermore, the omnibus tests indicated that AA position 74 was the most important within SE allele for ACPA development and thus might be a causal AA position to interact with CS. On the other hand, the effect of the interaction between CS and ACPA positivity was reported to be dependent on different SE subtypes.27 Although we did not observe difference between HLA-DRB1*04:05 and non-HLA-DRB1*04:05 SE possibly due to the different predominance of SE alleles among different ethnicities, other factors not explained by AA position 74 may also be of interest.

Previous studies with small cohort of Caucasian RA patients reported an association between CS and positivity40 or levels of RF in relation to SE.41 The former observed an almost independent association of CS with RF positivity, while the latter observed an additive effect of CS on RF levels with dose-dependent SE allele effect. Although the association of SE with RF levels might be different among different ethnicities partly due to the different SE allele predominance, our study suggests that CS can independently affect RF levels regardless of SE allele at least in Asian populations.

Taken together these findings our study further support that CS significantly affects ACPA and RF formation in RA patients, but in different manners especially in relation to SE alleles. One possible limitation of the present study is that ACPA and RF data used were those had been measured at the time of last visit instead of at onset. Thus, further studies enrolling more numbers of subjects from different races including non-RA subjects with complete sets of data including smoking information and autoantibody data are favourable to confirm the findings and further detailed analyses.

In conclusion, our study suggests that CS, especially at the time of onset, modify the risk of developing positive and higher levels of ACPA and RF in RA patients. CS may interact with SE and affect ACPA formation while the impact of CS on RF formation is independent of SE. Our findings suggest novel distinct mechanisms of RF and ACPA development in RA patients.

References

Footnotes

  • Handling editor Josef S Smolen

  • Contributorship Statement YI analysed and interpreted data and wrote the manuscript. KI, MH, KO, MT, HI, AT, HY and TM interpreted data and edited the manuscript. CT conceived the project, analysed and interpreted data, and wrote the manuscript.

  • Funding KI received scholarship donations from three pharmaceutical companies (Eisai, Chugai and Eli-Lilly). MH received scholarship donations from three pharmaceutical companies (Astellas, Bristol-Myers Squibb and Eisai). MT received scholarship donations from seven pharmaceutical companies (Pfizer Japan Inc., Astellas, Takeda Pharmaceutical, AbbVie GK Inc., Taisho-Toyama Pharmaceutical, Eisai and Asahi-Kasei). HI received scholarship donations from three pharmaceutical companies (Astellas, Bristol-Myers Squibb and Asahi-Kasei). AT received scholarship donations from two pharmaceutical companies (Eisai and Mitsubishi-Tanabe Pharma). TM received scholarship donations from thirteen pharmaceutical companies (Pfizer Japan Inc., Mitsubishi-Tanabe, Eisai, Astellas, AYUMI, Chugai, Daiichi Sankyo, Nippon Shinyaku, Takeda Pharmaceutical, Eli-Lilly, Asahi-Kasei Pharma, Nippon Kayaku and Sanofi).

  • Competing interests KI received speaker fee from fifteen pharmaceutical companies (Asahi-Kasei Pharma, Astellas Pharma, Abbvie GK Inc., AYUMI, Eisai, Otsuka, Kaken, Takeda, Mitsubishi-Tanabe, Chugai, Eli-Lilly, Bristol-Myers Squibb, Pfizer Japan Inc., Janssen and UCB Japan). MH, HI and MT belong to a department that has been financially supported by four pharmaceutical companies (Mitsubishi-Tanabe, Chugai, AYUMI and UCB Japan). TM received personal fees from fourteen pharmaceutical companies (Pfizer Japan Inc., Mitsubishi-Tanabe, Eisai, Astellas, AYUMI, Chugai, Daiichi Sankyo, Nippon Shinyaku, Takeda Pharmaceutical, Eli-Lilly, Asahi-Kasei Pharma, Sanofi, Bristol-Myers Squibb and GlaxoSmithKline). KURAMA cohort study is supported by grant from Daiichi Sankyo Co. Ltd. This study is conducted as investigator initiate study. These companies had no role in the design of the study, the collection or analysis of the data, the writing of the manuscript or decision to submit the manuscript for the publication. IORRA cohort study is supported by grant by twenty pharmaceutical companies (Daiichi Sankyo Co. Ltd., Mitsubishi-Tanabe, Chugai, Bristol-Myers Squibb, AYUMI, Astellas, Pfizer Japan Inc., Takeda Pharmaceutical, Eisai, Nippon Shinyaku, YL Biologics Ltd., AbbVie, Novartis Pharmaceutical K.K., Kaken Pharmaceutical Co., Ltd., UCB Japan, Ono Pharmaceutical Co., Ltd., Taisho Toyama Pharmaceutical Co., Ltd., Teijin Pharma, Torii Pharmaceutical Co., Ltd. and Nippon Boehringer Ingelheim Co., Ltd.). These sponsors were not involved in the: study design; collection, analysis and interpretation of data; writing of the paper; and/or decision to submit for publication.

  • Patient and public involvement statement This research was done without patient involvement. Patients were not invited to comment on the study design and were not consulted to develop patient relevant outcomes or interpret the results. Patients were not invited to contribute to the writing or editing of this document for readability or accuracy.

  • Patient consent for publication Not required.

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

  • Data availability statement Individual-level deidentified patient data, survey results, interview transcripts, statistical code and spreadsheets would be available from Chikashi Terao (ORCID 0000-0002-6452-4095) at any time only on reasonable request. There is no additional information available.