You are here

  • Home
  • Archive
  • Volume 10, Issue 2
  • Polygenic Risk Score Predicts Earlier-Onset Adult Systemic Lupus Erythematosus and First-Year Renal Diseases in a Taiwanese Cohort

Article Text

Article menu

Download PDFPDF

Lupus
Original research
Polygenic Risk Score Predicts Earlier-Onset Adult Systemic Lupus Erythematosus and First-Year Renal Diseases in a Taiwanese Cohort
  1. Yen-Ju Chen1,2,3,
  2. Tzu-Hung Hsiao1,4,5,
  3. Ying-Cheng Lin6,
  4. Wen-Juei Jeng7,8,
  5. Chien-Lin Mao1,
  6. Chia-Yi Wei1,
  7. Yi-Chung Hsieh7,8,
  8. Chih-Jen Huang9,
  9. Mei-Hung Pan9,
  10. I-Chieh Chen1,
  11. Ching-Heng Lin1,10,11,12,
  12. Yi-Ming Chen1,2,13 and
  13. Hwai-I Yang3,9,14,15
  1. 1Department of Medical Research, Taichung Veterans General Hospital, Taichung, Taiwan
  2. 2Division of Allergy, Immunology and Rheumatology, Department of Internal Medicine, Taichung Veterans General Hospital, Taichung, Taiwan
  3. 3Institute of Clinical Medicine, National Yang Ming Chiao Tung University, Taipei, Taiwan
  4. 4Department of Public Health, College of Medicine, Fu Jen Catholic University, New Taipei City, Taiwan
  5. 5Institute of Genomics and Bioinformatics, National Chung Hsing University, Taichung, Taiwan
  6. 6Division of Gastroenterology and Hepatology, Department of Internal Medicine, Taichung Veterans General Hospital, Taichung, Taiwan
  7. 7Division of Hepatogastroenterology, Chang Gung Memorial Hospital, Taoyuan City, Taiwan
  8. 8College of Medicine, Chang Gung University, Taoyuan City, Taiwan
  9. 9Genomics Research Center, Academia Sinica, Taipei, Taiwan
  10. 10Department of Health Care Management, National Taipei University of Nursing and Health Sciences, Taipei, Taiwan
  11. 11Department of Industrial Engineering and Enterprise Information, Tunghai University, Taichung, Taiwan
  12. 12Institute of Public Health and Community Medicine Research Center, National Yang Ming Chiao Tung University, Taipei, Taiwan
  13. 13Institute of Biomedical Science and Rong Hsing Research Center for Translational Medicine, National Chung Hsing University, Taichung, Taiwan
  14. 14Graduate Institute of Medicine, College of Medicine, Kaohsiung Medical University, Kaohsiung, Taiwan
  15. 15Biomedical Translation Research Center, Academia Sinica, Taipei, Taiwan
  1. Correspondence to Dr Hwai-I Yang; hiyang{at}gate.sinica.edu.tw; Dr Yi-Ming Chen; ymchen1{at}vghtc.gov.tw

Abstract

Objectives This study aimed to develop a predictive model using polygenic risk score (PRS) to forecast renal outcomes for adult systemic lupus erythematosus (SLE) in a Taiwanese population.

Methods Patients with SLE (n=2782) and matched non-SLE controls (n=11 128) were genotyped using Genome-Wide TWB 2.0 single-nucleotide polymorphism (SNP) array. PRS models (C+T, LDpred2, Lassosum, PRSice-2, PRS-continuous shrinkage (CS)) were constructed for predicting SLE susceptibility. Logistic regression was assessed for C+T-based PRS association with renal involvement in patients with SLE.

Results In the training set, C+T-based SLE-PRS, only incorporating 27 SNPs, outperformed other models with area under the curve (AUC) values of 0.629, surpassing Lassosum (AUC=0.621), PRSice-2 (AUC=0.615), LDpred2 (AUC=0.609) and PRS-CS (AUC=0.602). Additionally, C+T-based SLE-PRS demonstrated consistent predictive capacity in the testing set (AUC=0.620). Individuals in the highest quartile exhibited earlier SLE onset (39.06 vs 44.22 years, p<0.01), higher Systemic Lupus Erythematosus Disease Activity Index scores (3.00 vs 2.37, p=0.04), elevated risks of renal involvement within the first year of SLE diagnosis, including WHO class III–IV lupus nephritis (OR 2.36, 95% CI 1.47 to 3.80, p<0.01), estimated glomerular filtration rate <60 mL/min/1.73m2 (OR 1.49, 95% CI 1.18 to 1.89, p<0.01) and urine protein-to-creatinine ratio >150 mg/day (OR 2.07, 95% CI 1.49 to 2.89, p<0.01), along with increased seropositivity risks, compared with those in the lowest quartile. Furthermore, among patients with SLE with onset before 50 years, the highest PRS quartile was significantly associated with more serious renal diseases within the first year of SLE diagnosis.

Conclusions PRS of SLE is associated with earlier onset, renal involvement within the first year of SLE diagnosis and seropositivity in Taiwanese patients. Integrating PRS with clinical decision-making may enhance lupus nephritis screening and early treatment to improve renal outcomes in patients with SLE.

  • Systemic Lupus Erythematosus
  • Polymorphism, Genetic
  • Lupus Nephritis

Data availability statement

Data are available on reasonable request.

http://creativecommons.org/licenses/by-nc/4.0/

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.

https://doi.org/10.1136/rmdopen-2023-003293

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.

    WHAT IS ALREADY KNOWN ON THIS TOPIC

    • Previous reports have identified some predictive factors for progression to lupus nephritis (LN), such as high serum creatinine at onset of systemic lupus erythematosus (SLE) and hypocomplementaemia, but there is still a lack of indicators for prediction of early renal diseases in patients with SLE.

    WHAT THIS STUDY ADDS

    • The polygenic risk score (PRS) constructed using clumping+thresholding algorithm outperformed other prediction models in predicting SLE development in the Taiwanese population.

    • PRS of SLE was associated with earlier onset, autoantibody seropositivity and renal involvement within the first year of SLE diagnosis in Taiwanese patients with SLE.

    • PRS of SLE had better predictability for renal involvement in younger-onset patients with SLE (<50 years) in the Taiwanese population, compared with late-onset patients with SLE (50 years).

    HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

    • Integrating the PRS with clinical decision-making may enhance LN screening and facilitate early treatment to improve renal outcomes in patients with SLE. For patients with SLE with high SLE-PRS score, close monitor and follow-up were recommended.

    Introduction

    Systemic lupus erythematosus (SLE) is a systemic autoimmune disease with variable manifestations, leading to delays in diagnosis and treatment. Despite advances in biotechnology that allow for early identification and timely management of SLE, the mortality rate remains higher in patients with SLE as compared with matched controls.1–4 One of the main causes of death in patients with SLE is uncontrolled lupus nephritis (LN), especially in Asian populations.5 The pathogenesis of LN involves the loss of tolerance to nuclear autoantigens, production of proinflammatory cytokines, activation of complement systems and immune complex-induced intrarenal and extrarenal inflammation.6 7 Previous reports have identified some predictive factors for progression to LN, including high serum creatinine at onset of SLE and hypocomplementaemia, while predictors for end-stage renal disease (ESRD) in LN patients included poor early treatment response, class III/IV/VI or high chronicity index by histopathological findings and male gender.8–10 Nevertheless, there is still a lack of indicators for early prediction of renal impairment in patients with SLE. Baseline clinical and pathological data are currently insufficient to provide useful information on the future development of LN and medical comorbidities in patients with SLE.

    With the advancement of genetic technology, an increasing number of studies have demonstrated convincing evidence of genetic susceptibility to SLE.11 12 Earlier genome-wide association studies (GWASs) have identified numerous genetic variants associated with SLE, including the human leucocyte antigen alleles, genes of complement components and genes associated with lymphocyte development, differentiation and signalling.13 Furthermore, different subsets of risk alleles may correspond to the development of different phenotypes in SLE.14 For example, variants of Signal Transducer and Activator of Transcription 4 (STAT4) and B Cell Scaffold Protein With Ankyrin Repeats 1 have been found to be associated with LN and severe renal impairments in patients with SLE.15 16 Notably, evidence suggests that SLE is a polygenic disease, and the contribution of each risk allele to phenotypes may vary. Therefore, a comprehensive assessment of the impact of genotypes on phenotypes is necessary for patients with SLE to facilitate personalised medicine.

    Polygenic risk score (PRS) is a method used to calculate the weighted cumulative effect of multiple risk alleles, which can be further explored to understand its association with specific phenotypes. Given the inheritable characteristics of SLE, it is reasonable to expect that an exploration of the genetic contribution of PRS construction to LN and medical comorbidities may yield important information. Moreover, precise prediction models may facilitate early detection and interventions to improve clinical outcomes of SLE. Reid et al demonstrated that patients with SLE with high PRS scores in Western populations have a higher risk of ESRD (OR 5.58, 95% CI 1.50 to 20.79) and proliferative nephritis (OR 2.42, 95% CI 1.30 to 4.49), as well as a higher prevalence of damage accrual (OR 1.47, 95% CI 1.06 to 2.04).17 However, since serum creatinine and urine protein levels were not recorded, it may lead to an underestimation of early onset of renal diseases in patients with SLE. Moreover, higher prevalence and earlier onset of SLE have been noted in Asian populations than in Caucasians, implying different genetic architectures exist among ethnicities.18 19 Currently, only a few studies have examined the association of weighted genetic risks and SLE in Asian populations, and no PRS models for LN developed using Asian populations have been reported.20

    In this study, we aimed to develop a tailored predictive PRS model for the Taiwanese population, and to examine the associations of PRS with renal diseases within the first year of SLE diagnosis in patients with SLE.

    Material and methods

    Enrolment of participants and identification of SLE cases

    Between July 2019 and May 2022, we enrolled participants aged ≥20 years from outpatient departments in a tertiary medical centre in Taiwan, in partnership with the Taiwan Precision Medicine Initiative project supervised by Academia Sinica, Taiwan. Subjects with active leukaemia, patients receiving blood transfusions within the past 6 months and patients with malignancy who received chemotherapy or radiotherapy within 1 year were excluded. Detailed information of participants, including medical records, physical examinations and blood tests, was collected and all participants underwent genotyping using the Affymetrix Genome-Wide TWB 2.0 single-nucleotide polymorphism (SNP) Array. SLE diagnosis was based on either the 1997 American College of Rheumatology (ACR) classification criteria or the 2012 Systemic Lupus International Collaborating Clinics (SLICC) classification criteria.21 22 Study cases were identified using the International Classification of Diseases-Ninth Revision-Clinical Modification (ICD-9-CM) code 710.0 at least three times in outpatient visits or once in a hospitalisation. We further confirmed the cases by the issuance of catastrophic illness certificate for SLE. Patients who had major illnesses such as cancers and specific autoimmune diseases were registered in the National Health Insurance programme in Taiwan. A catastrophic illness certificate is issued after a thoroughly review of medical records by at least two qualified rheumatologists. Controls were selected after excluding immune-related disorders, matching with SLE cases for age and gender, and controlling for ethnicity at a ratio of 1:4.

    Genotyping and quality controls

    The blood samples of the participants were obtained for DNA extraction and were further genotyped using the Affymetrix Genome-Wide TWB 2.0 SNP Array (Affymetrix, Santa Clara, California, USA).23 The Affymetrix Genome-Wide TWB 2.0 SNP chip was specifically designed for Taiwan’s Han Chinese population and consisted of 714 431 SNPs. Affymetrix Power Tools was used for processing and analysing array data, and quality control procedures were carried out to exclude markers related to sex hormones, those with minor allele frequency <0.01 or a genotype missing rate of >5%, and those that failed the Hardy-Weinberg equilibrium tests with a p<1.0×10−5. A total of 591 048 SNPs were retained for further analysis after quality control.

    GWAS and construction of PRS predictive models

    GWASs were performed in training sets to determine the putative susceptibility variants for SLE with genome-wide significant p<5×10−8. PRS was then quantified by calculating the sum of the effect alleles in a subject, weighted by the estimated effect size of corresponding genotypes from our GWAS result on the phenotype.24 Five methods for deriving PRS, including clumping+thresholding (C+T), PRSice-2, LDpred2, Lassosum and PRS-continuous shrinkage (CS) models,25–29 were used to develop five candidate PRS in the training set, and they were named after their derivation methods. These five methods can be mainly classified into two analytical steps for PRS scoring: (1) methods that clump SNPs from GWAS and sum their effects, such as standard C+T method considering the p value thresholding,25 and PRSice-2 method,26 which is an automated PRS analysis procedure that improves on the original PRSice method by providing empirically associated p values to avoid inflation due to overfitting; (2) methods that consider all SNPs from GWAS and their linkage disequilibrium (LD) structure while applying traditional shrinkage techniques, including LDpred-2, Lassosum and PRS-CS. The LDpred-2 algorithm is a PRS programme that uses a Bayesian shrinkage in its PRS calculation.27 The Lassosum algorithm uses penalised regression, specifically the least absolute shrinkage and selection operator) technique, for polygenic risk scoring,28 while the PRS-CS algorithm uses a Bayesian regression with CS prior to computing SNP effect sizes on large-scale data.29 The area under the curve (AUC) for SLE susceptibility, based on each model, was evaluated to determine the prediction accuracies of the PRS models. Based on the predictive performance of the PRS, C+T model was further employed for disease risk and clinical outcome prediction in a Taiwanese cohort of patients with SLE. To validate the results and evaluate the clinical impact of PRS, all patients with SLE were ranked from the lowest PRS score to the highest score, and the patients with SLE were further divided into four groups according to their individualised PRS scores. The four groups were as follows: the lowest quartile, the second quartile, the third quartile and the highest quartile. The lowest quartile group included one-quarter of patients with SLE with the lowest PRS, and the highest quartile group included one quarter of patients with SLE with the highest PRS. Associations of PRS with clinical characteristics and renal outcomes in patients with SLE were examined across the four groups using logistic regression analysis, and ORs and 95% CIs were estimated.

    Definition of covariables

    Information on comorbidities was identified by the ICD-9-CM codes recorded three times or more during outpatient visits or at least once in a hospitalisation within the first year of SLE diagnosis, or associated medication use during this period for at least 3 months, including hypertension (ICD-9-CM codes 401–405), diabetes mellitus (ICD-9-CM codes 250) and hyperlipidaemia (ICD-9-CM codes 272). Information regarding the status of specific medical comorbidities was included if coded three times or more during outpatient visits or at least once in a hospitalisation following the diagnosis of SLE. This included coronary artery disease (ICD-9-CM codes 411–413, 414.00, 414.01), chronic kidney disease (CKD, ICD-9-CM codes 585), cerebrovascular accident (CVA, ICD-9-CM codes 433–438), interstitial lung disease (ILD, ICD-9-CM codes 515), pulmonary hypertension (PH, ICD-9-CM codes 416.0, 416.8) and osteoporosis (ICD-9-CM codes 733). The Systemic Lupus Erythematosus Disease Activity Index (SLEDAI) for disease activity evaluations were collected within 1 year of SLE diagnosis.30 Information regarding the usage of antirheumatic medications, from the point of SLE diagnosis until the end of the follow-up period, was also collected. This included corticosteroids, hydroxychloroquine, azathioprine, methotrexate, cyclosporine, mycophenolate and cyclophosphamide. Renal pathology31 and biochemical data associated with renal diseases within the first year of SLE diagnosis, such as serum creatinine level and spot urine protein-to-creatinine ratio (UPCR), were gathered in patients with SLE. Abnormal UPCR was defined as greater than or equal to 150 mg per day. Severe renal disease was defined as persistent estimated glomerular filtration rate (eGFR) less than 60 mL/min/1.73 m2 for at least 6 months after the first year of SLE diagnosis. In addition, immunological markers, including antinuclear antibodies (ANA), anti-dsDNA antibodies, serum C3, serum C4, anti-Smith antibodies, anti-RNP antibodies, lupus anticoagulant, anticardiolipin antibodies, anti-B2GP1 antibodies and results of direct Coombs tests, were also collected within 1 year of SLE diagnosis. In addition, a subgroup analysis by onset age at SLE diagnosis was also performed to determine the extent of genetic contributions between younger-onset and late-onset SLE. Those diagnosed with SLE before 50 years were classified as younger onset, and those diagnosed at or after 50 years were deemed to have late-onset SLE, as previously described.32

    Patient and public involvement

    Patients or the public were not involved in the design, or conduct, or reporting, or dissemination plans of our research.

    Statistical analysis

    The demographic data are presented as mean±SD for continuous variables and number (per cent) for categorical variables. A Student’s t-test or analysis of variance for continuous variables and χ2 test for categorical variables were conducted. PLINK and R software (V.4.0.3) were used for PRS construction and the discriminatory capability of PRS to predict the development of SLE was assessed and expressed as AUC. The associations between PRS and clinical variables or renal outcomes were examined by multivariable-adjusted logistic regression analysis and ORs and 95% CIs were estimated. The Cochran-Armitage trend test was used for trend analysis in age subgroups. The data were analysed using SAS V.9.4 (SAS Institute). Statistical significance was set at p<0.05.

    Results

    Baseline characteristics of the participants

    There were 58 091 participants enrolled in the study, with 3023 patients with SLE identified and 54 986 non-SLE controls after excluding those who failed the quality control procedure (n=82) (figure 1). SLE cases without an available diagnosis date by medical records or those with missing SNPs were further excluded. As a result, 2782 patients with SLE were included in the final analysis. The individuals were randomly divided into two groups: the training set included 1782 patients with SLE and 7128 non-SLE controls, while the testing set comprised 1000 patients with SLE and 4000 non-SLE controls after matching at a 1:4 ratio.

    Figure 1
    Figure 1

    Overview of study design. SLE, systemic lupus erythematosus; SNP, single-nucleotide polymorphism.

    GWAS for SLE and PRS predictive models for the Taiwanese population

    The Manhattan plot of SLE in the Taiwanese population is shown in online supplemental figure 1. In total, 203 SNPs were found to be associated with SLE in the Taiwanese population using genome-wide association analysis. In a previous investigation, the most significant SNP associated with Taiwanese patients with SLE was GTF2I/rs117026326 (p=4.26×10−51), which was especially prevalent in Asian populations. Other significant SLE-associated variants in the Taiwanese population are summarised in online supplemental table 1.

    We developed PRS according to five algorithms and conducted ROC analysis of these PRS for SLE outcome. The results showed that the AUC of PRS calculated by C+T, Lassosum, PRSice-2, LDpred2 and PRS-CS models in the training set were 0.629, 0.621, 0.615, 0.609, 0.602, respectively, and in the testing set were 0.620, 0.612, 0.609, 0.646, 0.597, respectively (figure 2). Given its superior prediction capacity in the training set and strong consistency in the testing group, the PRS constructed using the C+T algorithm (denoted as the SLE-PRS) was used for subsequent analyses.

    Figure 2
    Figure 2

    Prediction accuracy of the PRS of SLE developed using various algorithms in the Taiwanese population in the (A) training and (B) testing dataset. AUC, area under the curve; CS, continuous shrinkage; C+T, clumping+thresholdin; PRS, polygenic risk score; SLE, systemic lupus erythematosus.

    The association of SLE-PRS and clinical characteristics in the Taiwanese SLE population

    Table 1 shows the distributions of clinical characteristics of patients with SLE by SLE-PRS quartiles. The group in the highest quartile of SLE-PRS scores exhibited earlier SLE onset (the lowest vs the highest quartile: 44.22 vs 39.06 years, p<0.01), and the proportion of patients experiencing the onset of SLE before the age of 25 years was 12.14%, 13.07%, 13.09% and 17.74% in the lowest, second, third and the highest quartile, respectively (p=0.01). The patients with SLE in the highest quartile of SLE-PRS scores also had higher SLEDAI scores (the lowest vs the highest quartile: 2.37 vs 3.00, p=0.04), and a higher percentage of patients using corticosteroids (the lowest vs the highest quartile: 80.64% vs 89.99%, p<0.01), hydroxychloroquine (the lowest vs the highest quartile: 83.53% vs 94.28%, p<0.01), azathioprine (the lowest vs the highest quartile: 30.78% vs 54.65%, p<0.01), mycophenolate (the lowest vs the highest quartile: 11.71% vs 27.32%, p<0.01), cyclosporine (the lowest vs the highest quartile: 14.45% vs 22.60%, p<0.01) and cyclophosphamide (the lowest vs the highest quartile: 10.40% vs 19.74%, p<0.01). However, they had lower percentages of DM (the lowest vs the highest quartile: 9.83% vs 9.01%, p=0.04). After adjusting for potential confounders, significantly higher SLEDAI scores and increased need for use of several immunosuppressive agents were observed in the highest SLE-PRS quartile as compared with the remaining quartiles (online supplemental table 2).

    View this table:
    Table 1

    Demographics of patients with SLE by quartiles of SLE-PRS scores

    The association between SLE-PRS and renal involvement within the first year of SLE diagnosis

    Renal involvement is one of the most important prognostic factors for patients with SLE and timely treatment is crucial in preventing ESRD and renal transplantation. A rising percentage of UPCR>150 mg/day, ranging from 12.28% to 26.75% and decreasing serum albumin levels, ranging from 4.08 ml/dL to 3.91 mL/dL, were observed from the lowest quartile to the highest quartile (table 1). The ORs of eGFR<60 mL/min/1.73 m2 in the highest quartile were 1.49 (95% CI 1.18 to 1.89, p<0.01), the ORs of abnormal UPCR in the third quartile and the highest quartile were 1.46 (95% CI 1.04 to 2.06, p=0.03) and 2.07 (95% CI 1.49 to 2.89, p<0.01), respectively, in patients with SLE, compared with the lowest quartile (figure 3A and online supplemental table 2). Additionally, increasing risks of decline of serum albumin levels were also observed in patients with SLE in the third quartile and the highest quartile (OR 0.73, 95% CI 0.55 to 0.96, p=0.02; OR 0.63, 95% CI 0.48 to 0.83, p<0.01, respectively), as compared with those in the lowest quartile. Moreover, after adjusting for the duration of observation, ORs of CKD was 1.47 (95% CI 1.14 to 1.92, p<0.01), 1.72 (95% CI 1.33 to 2.22, p<0.01) and 2.69 (95% CI 2.08 to 3.48, p<0.01) in the second, the third and the highest quartile, respectively, compared with the lowest quartile.

    Figure 3
    Figure 3

    Associations of SLE-PRS and autoantibodies as well as early onset of renal diseases. Higher PRS was associated with (A) indicators for renal diseases within first years of SLE diagnosis and (B) seropositivity of autoantibodies in patients with SLE in the Taiwanese population. ANA, antinuclear antibody; eGFR, estimated glomerular filtration rate; PRS, polygenic risk score; SLE, systemic lupus erythematosus.

    An analysis was conducted to explore the relationship between renal histological data and SLE-PRS. Of the 236 patients who underwent renal biopsy within the first year of SLE diagnosis, the distribution across WHO classes was as follows: class I–II (n=12), class III–IV (n=156) and class V (n=68); no patients were categorised under WHO class VI. A trend was observed wherein the percentage of patients with WHO class III–IV LN increased from 11.02% in the lowest PRS quartile to 26.69% in the highest quartile. Similarly, the percentage of patients with WHO class V LN ranged from 2.54% in the lowest quartile to 8.9% in the highest quartile (table 1). In the highest PRS quartile, the OR for WHO class III–IV LN was 2.36 (95% CI 1.47 to 3.80, p<0.01), accompanied by a decrease in the chronicity score (OR 0.80, 95% CI 0.65 to 0.98, p=0.03) when compared with the lowest quartile (online supplemental table 2). Additionally, elevated risks for WHO Class V LN were observed in the second (OR 2.96, 95% CI 1.17 to 7.53, p=0.02), third (OR 3.87, 95% CI 1.56 to 9.57, p<0.01) and highest PRS quartiles (OR 3.47, 95% CI 1.39 to 8.69, p<0.01), relative to the lowest quartile.

    A separate subgroup analysis of patients with LN and the relationship of PRS with the renal parameters at 1 year were further explored. In total, there were 187 patients with UPCR over 150 mg/day. Decreasing renal survival at 6 months was found especially in LN patient in the highest and in the third quartiles of SLE-PRS scores (online supplemental figure 2). However, due to the lack of data on sequential treatment regimens, duration of follow-up, and the number of renal flares, we could not adjust the observation of renal survival in patients with LN for the above variables.

    The association between SLE-PRS and autoantibodies

    The proportions of positive ANA, positive anti-dsDNA antibody and hypocomplementaemia in terms of serum C3 and C4 levels increased gradually from the lowest quartile to the highest quartile of SLE-PRS scores (table 1). Figure 3B and online supplemental table 2 show that ORs were significantly elevated with increasing quartiles of SLE-PRS score for ANA positivity (the third quartile vs the lowest quartile: OR 1.39, 95% CI 1.01 to 1.90, p=0.04; the highest vs the lowest quartile: OR 1.59, 95% CI 1.15 to 2.20, p<0.01), anti-dsDNA antibody positivity (the second quartile vs the lowest quartile: OR 1.49, 95% CI 1.15 to 1.94, p<0.01; the third quartile vs the lowest quartile: OR 2.06, 95% CI 1.58 to 2.68, p<0.01; the highest vs the lowest quartile: OR 3.11, 95% CI 2.35 to 4.11, p<0.01) and hypocomplementaemia (low C3: the third quartile vs the lowest quartile: OR 1.66, 95% CI 1.28 to 2.16, p<0.01; the highest vs the lowest quartile: OR 2.65, 95% CI 2.03 to 3.47, p<0.01); (low C4: the second quartile vs the lowest quartile: OR 1.47, 95% CI 1.09 to 2.00, p=0.01; the third quartile vs the lowest quartile: OR 2.05, 95% CI 1.54 to 2.74, p<0.01; the highest vs the lowest quartile: OR 2.44, 95% CI 1.82 to 3.27, p<0.01) when compared with the lowest quartile group.

    Predictability of SLE-PRS by SLE onset age

    We then assessed the impact of genotypes on phenotypes among disease onset ages to evaluate if there were different genetic contributions between younger-onset (SLE diagnosed age <50 years) and late-onset (SLE diagnosed age ≥50 years) SLE. The median (IQR) SLE-PRS scores of younger-onset SLE and late-onset SLE were 2.18 (2.16) and 1.74 (2.14), respectively (p<0.01). Online supplemental table 3 shows the detailed median SLE-PRS scores of younger-onset SLE and late-onset SLE in the lowest, the second, the third and the highest quartile groups (younger-onset SLE: 0.19 (1.02), 1.52 (0.52), 2.54 (0.53) and 3.85 (1.05), respectively; late-onset SLE: 0.21 (0.97), 1.49 (0.57), 2.45 (0.55) and 3.68 (0.81), respectively). In the younger-onset SLE patient group, there was a trend in which the proportion of ANA positivity, dsDNA positivity and hypocomplementaemia increased from patients with SLE in the lowest quartile to those in the highest quartile of the SLE-PRS scores (the lowest to the highest quartile: ANA: 66.67% to 79.05%, p for trend<0.01; dsDNA: 54.50% to 83.73%, p for trend<0.01; low C3: 46.10% to 69.74%, p for trend<0.01; low C4: 26.13% to 46.92%, p for trend<0.01) (figure 4, online supplemental tables 3 and 4). Moreover, in the younger-onset SLE group, there was a trend towards poor renal outcomes from low SLE-PRS scores to high scores, including an increasing proportion of CKD, eGFR<60 mL/min/1.73 m2, and UPCR>150 mg/day, from 3.31% to 6.65% (p for trend=0.03), 23.88% to 32.13% (p for trend=0.03) and 17.26% to 31.75% (p for trend<0.01), respectively.

    Figure 4
    Figure 4

    Cochran-Armitage trend test between SLE-PRS quartile and seropositivity of autoantibodies, early onset of renal diseases and comorbidities stratified by age at SLE onset. There was a trend showing (A) renal diseases within first years of SLE diagnosis and renal damage and (B) seropositivity, were increased among patients with SLE in the lowest quartile compared with those in the highest quartile of the SLE-PRS scores, especially in the younger-onset SLE group. SLE-PRS might play a more important role in SLE disease course prediction in younger onset patients. *Cochran-Armitage trend test (two-tailed test) p<0.05. ANA, antinuclear antibody; CKD, chronic kidney disease; eGFR, estimated glomerular filtration rate; PRS, polygenic risk score; SLE, systemic lupus erythematosus.

    Multivariable-adjusted logistic regression analysis further revealed younger-onset patients with SLE in the highest quartile of SLE-PRS scores exhibited significantly higher risks of elevated dsDNA (OR 4.32, 95% CI 3.16 to 5.90, p<0.01), hypocomplementaemia (C3: OR 2.84, 95% CI 2.07 to 3.89, p<0.01; C4: OR 2.68, 95% CI 1.93 to 3.72, p<0.01) and renal diseases within the first year of SLE diagnosis including UPCR>150 mg/day (OR 2.30, 95% CI 1.67 to 3.15, p<0.01), WHO class III–IV (OR 2.43, 95% CI 1.48 to 4.02, p<0.01) and WHO class V (OR 3.04, 95% CI 1.21 to 7.67, p=0.02), but lower risks of chronicity score (OR 0.81, 95% CI 0.65 to 0.99, p=0.04) among the younger-onset SLE patient group in the highest quartile of SLE-PRS scores, as compared with those in the lowest quartile (table 2). Increasing risks of medical comorbidities with CKD (OR 3.15, 95% CI 2.29 to 4.32, p<0.01) and osteoporosis (OR 2.69, 95% CI 1.78 to 4.04, p<0.01) was also observed in the younger-onset SLE patient group in the highest quartile of SLE-PRS. In the late-onset SLE group, a high SLE-PRS score was only associated with an increased risk of low serum C3 levels (OR 1.71, 95% CI 1.00 to 2.91, p=0.04) and UPCR>150 mg/day (OR 2.79, 95% CI 1.32 to 5.86, p<0.01), as compared with a low score.

    View this table:
    Table 2

    Associations of SLE-PRS with comorbidities, and biochemical and immunological tests stratified by age of SLE onset

    Discussion

    In this study, we constructed a PRS specific to individuals of Taiwanese ancestry for the prediction of SLE outcome. Elevated SLE-PRS scores were associated with earlier onset of SLE, higher SLEDAI scores, and increased need of immunosuppressive agents. Additionally, high SLE-PRS scores were linked to renal involvement within the first year of SLE diagnosis, and seropositivity for ANA, anti-dsDNA antibody and hypocomplementaemia, particularly in younger-onset patients with SLE. Our findings have potential implications for applying PRS in screening programmes and clinical decision-making for LN. Moreover, identifying high-risk groups based on PRS can facilitate early treatment and improve renal outcomes for patients with SLE in the Taiwanese population.

    Consistent with previous reports, our study demonstrated similar predictive capability for SLE development with an AUC of 0.62. However, we observed higher SLE severity and prevalence of nephritis in Asian populations, implying different genetic architectures across ethnicities. One meta-analysis identified 113 susceptibility loci to SLE, including 46 novel genetic regions with genome-wide significance in 208 370 East Asians.33 Wang et al conducted a GWAS study in Asian populations and developed multiple PRS models for SLE among different ethnicities using 293 684 genetic variants. They found that the prediction power of the ancestry-matched algorithm for SLE was significantly better than that of the ancestry-mismatched algorithm (AUC 0.76 (0.74–0.78) vs 0.62 (0.60–0.64)). Another study by Khunsriraksakul et al explored PRS of SLE derived from multiancestry and multitrait GWAS data, and the results suggested that PRS from multiancestry might outperform that from isolated ancestry, which would improve the diagnostic accuracy of SLE.34 Nonetheless, there may be variations in gene penetrance and protein levels among subjects of diverse genetic backgrounds.35 In addition, incorporating a large number of risk alleles from multiple ancestries into PRS construction may pose challenges in clinical implementation. In the present study, we selected 27 SNPs from Taiwanese GWAS data with genome-wide significance to establish a PRS for SLE susceptibility, which demonstrated satisfactory AUC and excellent clinical correlation.

    Our findings showed that high SLE-PRS scores were indicative of renal involvement within the first year of SLE diagnosis, including abnormal UPCR, hypoalbuminaemia and a trend for low eGFR. A study by Reid et al established a relationship between PRS scores of SLE and nephritis, with an elevated PRS score being associated with later stages of CKD, the development of ESRD, and earlier onset of ESRD.17 Chen et al also found a remarkable positive correlation between renal dysfunction and PRS scores.36 Nevertheless, previous studies have primarily focused on the association of PRS and well-defined LN, and the relationship with early onset of nephritis remained unclear. Autoimmunity gradually emerged before the development of clinical SLE, and renal inflammation in SLE could also be a progressive process.37 In our study, we identified candidate indicators of renal involvement within the first year of SLE diagnosis and further explored their positive association with PRS of SLE. Notably, the predictive capability of SLE-PRS is confined to early-onset renal disease observed within the first year after an SLE diagnosis, and it does not encompass long-term renal outcomes. By incorporating PRS and the indicators, high-risk groups for early onset of LN can be identified, enabling prompt therapies for patients with SLE. Indeed, in the past decade, the widespread use of immunosuppressive agents and the introduction of biological agents have greatly reduced renal inflammation and injury of LN, along with progression to ESRD in patients with SLE. Advances in genetic research on SLE could further promote the development of novel targeted therapies for SLE and LN to improve renal outcomes in patients with SLE, which would represent a significant stride toward precision medicine and individualised management. Regarding the cost-effectiveness of this approach, while our study did not directly assess this aspect, we hypothesised that the early identification of high-risk patients could lead to more targeted treatment, thereby potentially reducing long-term healthcare costs related to renal complications.

    The age of SLE onset differentially affects the initial presentation and clinical course, but controversies remain with regard to its impact on long-term outcomes and vital organ damage. Tomic-Lucic et al found higher frequencies of LN (46.6% vs 6.6%, p=0.006) and higher doses of corticosteroids (16.13 mg vs 11.76 mg, p=0.006) and cyclophosphamide (8.20 g vs6.26g, p=0.001) in the earlier-onset SLE group compared with late-onset patients with SLE.38 Sousa et al also suggested a greater proportion of renal involvement, higher disease activities and more patients using cyclophosphamide and mycophenolate mofetil in childhood-onset patients with SLE as compared with their counterparts.39 The higher prevalence of renal impairments and need for immunosuppressive agents support the notion that younger age at disease onset is associated with more severe presentation in patients with SLE. Our study demonstrated a superior predictive power for early onset of LN and seropositivity in younger-onset patients with SLE compared with the late-onset groups. These findings suggest that the prognostic disparities between younger-onset and late-onset SLE may extend beyond mere age or hormonal influences, implicating genetic factors as a potential determinant. Previous research has identified a specific SNP, STAT4 rs7574865, as being associated with an earlier age of SLE diagnosis.40 Beyond individual SNPs, cumulative risk alleles also had strong association with age at SLE diagnosis (p=9×10−7), as reported by Taylor et al.41 Our PRS prediction model provides an means for early risk stratification, particularly in younger-onset patients with SLE. Early diagnosis and timely treatment are essential for individuals predisposed to high disease severity, such as patients experiencing SLE onset at an earlier age.

    This study has several strengths. First, this study explored the impact of SLE-PRS on phenotypes with a large sample size, which allowed sufficient power for statistical analysis. In addition, the Affymetrix Genome-Wide TWB 2.0 SNP array is a validated chip with a rigorous quality control procedure, making it a valuable tool for genotyping East Asians. This approach facilitated the elucidation of the effects of tailor-made hotspot genes in clinical practice.

    However, there were some limitations in this study. First, the study was conducted in a single medical institution. Although we performed internal validation for improving the prediction accuracy of PRS in patients with SLE, external validation of PRS is warranted in an independent cohort in the future. Second, the participants were all Taiwanese, and thus, the results might not be generalisable to other ethnicities. Since the genetic backgrounds of Taiwanese are closer to those of East Asians than to Westerners, further genetic studies on other Asian populations should be conducted. Third, because we enrolled the participants in a tertiary medical centre, some patients had been diagnosed with SLE at other facilities prior to their inclusion in our study. As a result, we may not have fully captured the initial disease state at the time of SLE diagnosis. Future research could benefit from linking our cohort to a more comprehensive nationwide database, such as Taiwan’s National Health Insurance Research Database, to mitigate this limitation. Fourth, although information related to medical comorbidities was collected, but there was a lack of information of organ damage such as SLICC/ACR damage index score,42 which was a measure of morbidity, describing the accumulation of damage in patients with SLE, and was widely used for organ damage evaluation. Fifth, some factors associated with renal survival in patients with LN—such as sequential treatment regimens, duration of follow-up and number of renal flares and certain risk factors of SLE such as family history of SLE or other autoimmune disorders, sun exposure and viral infection—were not recorded in the study. Consequently, we could not totally exclude potential confounders in the present study.

    The study provides a valuable contribution to the establishment of a Taiwanese ancestry-specific PRS and its impact on phenotypes, particularly renal outcomes. This information may be helpful in developing a risk stratification strategy for patients with SLE based on their PRS scores. Identifying patients with SLE and nephritis as early as possible is crucial in terms of improving renal outcomes as well as reducing irreversible damage. Based on the results of our study, we recommend vigilant monitoring of renal function, urinary protein and serum albumin levels in patients with SLE identified with high PRS scores at the time of diagnosis, to enable early detection and management of LN. In addition, high PRS scores are associated with higher disease activities in SLE, and regular outpatient visits and enhancing medication adherence should be encouraged for high-risk groups. For individuals with high PRS scores who do not have SLE currently, shared decision-making should be considered to determine whether to conduct regular screening for SLE, and to avoid environmental triggers of lupus, such as sun exposure, oestrogen use and lupus-inducing medications, in order to prevent or delay the development of SLE.

    Conclusion

    PRS of SLE is associated with earlier onset, renal involvement within the first year of SLE diagnosis and seropositivity of autoantibodies in patients with SLE in the Taiwanese population. The PRS emerges as a promising tool for enhancing clinical decision-making in SLE. By calculating PRS scores at the time of diagnosis, clinicians can better tailor monitoring and treatment strategies. This is particularly relevant for patients with high PRS scores and younger onset, where early intervention with immunosuppressive agents targeting renal inflammation could not only mitigate adverse renal outcomes but also potentially reduce long-term healthcare costs. Prospective studies are warranted to determine the impact of adding PRS to clinical predictors on the prognosis of SLE outcomes, particularly the long-term outcomes in patients with SLE.

    Data availability statement

    Data are available on reasonable request.

    Ethics statements

    Patient consent for publication

    Ethics approval

    All participants provided written informed consent and the study was approved by the ethics committee of Taichung Veterans General Hospital’s Institutional Review Board (IRB No. SF19153A).

    Acknowledgments

    We would like to thank all of the participants and investigators from the Taiwan Precision Medicine Initiative and the biostatistics assistants at the Academia Sinica, and Department of Medical Research, Taichung Veterans General Hospital for assisting with the statistical analysis.The authors sincerely appreciate the assistance of the Center for Translational Medicine of Taichung Veterans General Hospital.

    References

    1. ↵
      1. Tselios K,
      2. Gladman DD,
      3. Sheane BJ, et al
      . All-cause, cause-specific and age-specific standardised mortality ratios of patients with systemic lupus erythematosus in Ontario, Canada over 43 years (1971-2013). Ann Rheum Dis 2019;78:802–6. doi:10.1136/annrheumdis-2018-214802
      OpenUrlAbstract/FREE Full Text
    2. ↵
      1. Rees F,
      2. Doherty M,
      3. Grainge MJ, et al
      . Mortality in systemic lupus erythematosus in the United Kingdom 1999-2012. Rheumatology (Oxford) 2016;55:854–60. doi:10.1093/rheumatology/kev424
      OpenUrlCrossRefPubMed
    3. ↵
      1. Bernatsky S,
      2. Boivin J-F,
      3. Joseph L, et al
      . Mortality in systemic lupus erythematosus. Arthritis Rheum 2006;54:2550–7. doi:10.1002/art.21955
      OpenUrlCrossRefPubMedWeb of Science
    4. ↵
      1. Chen Y-M,
      2. Lin C-H,
      3. Chen H-H, et al
      . Onset age affects mortality and renal outcome of female systemic lupus erythematosus patients: a nationwide population-based study in Taiwan. Rheumatology (Oxford) 2014;53:180–5. doi:10.1093/rheumatology/ket330
      OpenUrlCrossRefPubMed
    5. ↵
      1. Yap DYH,
      2. Tang CSO,
      3. Ma MKM, et al
      . Survival analysis and causes of mortality in patients with lupus nephritis. Nephrol Dial Transplant 2012;27:3248–54. doi:10.1093/ndt/gfs073
      OpenUrlCrossRefPubMedWeb of Science
    6. ↵
      1. Liu Y,
      2. Anders HJ
      . Lupus nephritis: from pathogenesis to targets for biologic treatment. Nephron Clin Pract 2014;128:224–31. doi:10.1159/000368581
      OpenUrlCrossRefPubMed
    7. ↵
      1. Sterner RM,
      2. Hartono SP,
      3. Grande JP
      . The pathogenesis of lupus nephritis. J Clin Cell Immunol 2014;5:205. doi:10.4172/2155-9899.1000205
    8. ↵
      1. Ocampo-Piraquive V,
      2. Nieto-Aristizábal I,
      3. Cañas CA, et al
      . Mortality in systemic lupus erythematosus: causes, predictors and interventions. Expert Rev Clin Immunol 2018;14:1043–53. doi:10.1080/1744666X.2018.1538789
      OpenUrlPubMed
    9. ↵
      1. Parodis I,
      2. Tamirou F,
      3. Houssiau FA
      . Prediction of prognosis and renal outcome in lupus nephritis. Lupus Sci Med 2020;7:e000389. doi:10.1136/lupus-2020-000389
    10. ↵
      1. Mahajan A,
      2. Amelio J,
      3. Gairy K, et al
      . Systemic lupus erythematosus, lupus nephritis and end-stage renal disease: a pragmatic review mapping disease severity and progression. Lupus 2020;29:1011–20. doi:10.1177/0961203320932219
      OpenUrlCrossRefPubMed
    11. ↵
      1. Järvinen P,
      2. Aho K
      . Twin studies in rheumatic diseases. Semin Arthritis Rheum 1994;24:19–28. doi:10.1016/0049-0172(94)90096-5
      OpenUrlCrossRefPubMedWeb of Science
    12. ↵
      1. Kelly JA,
      2. Moser KL,
      3. Harley JB
      . The genetics of systemic lupus erythematosus: putting the pieces together. Genes Immun 2002;3 Suppl 1:S71–85. doi:10.1038/sj.gene.6363885
      OpenUrlCrossRef
    13. ↵
      1. Kwon Y-C,
      2. Chun S,
      3. Kim K, et al
      . Update on the genetics of systemic lupus erythematosus: genome-wide association studies and beyond. Cells 2019;8:1180. doi:10.3390/cells8101180
    14. ↵
      1. Chen L,
      2. Morris DL,
      3. Vyse TJ
      . Genetic advances in systemic lupus erythematosus: an update. Curr Opin Rheumatol 2017;29:423–33. doi:10.1097/BOR.0000000000000411
      OpenUrlCrossRefPubMed
    15. ↵
      1. Almaani S,
      2. Meara A,
      3. Rovin BH
      . Update on lupus nephritis. Clin J Am Soc Nephrol 2017;12:825–35. doi:10.2215/CJN.05780616
      OpenUrlAbstract/FREE Full Text
    16. ↵
      1. Bolin K,
      2. Imgenberg-Kreuz J,
      3. Leonard D, et al
      . Variants in BANK1 are associated with lupus nephritis of European ancestry. Genes Immun 2021;22:194–202. doi:10.1038/s41435-021-00142-8
      OpenUrlCrossRefPubMed
    17. ↵
      1. Reid S,
      2. Alexsson A,
      3. Frodlund M, et al
      . High genetic risk score is associated with early disease onset, damage accrual and decreased survival in systemic lupus erythematosus. Ann Rheum Dis 2020;79:363–9. doi:10.1136/annrheumdis-2019-216227
      OpenUrlAbstract/FREE Full Text
    18. ↵
      1. Morais SA,
      2. Isenberg DA
      . A study of the influence of ethnicity on serology and clinical features in lupus. Lupus 2017;26:17–26. doi:10.1177/0961203316645204
      OpenUrlCrossRefPubMed
    19. ↵
      1. Morris DL,
      2. Sheng Y,
      3. Zhang Y, et al
      . Genome-wide association meta-analysis in Chinese and European individuals identifies ten new loci associated with systemic lupus erythematosus. Nat Genet 2016;48:940–6. doi:10.1038/ng.3603
      OpenUrlCrossRefPubMed
    20. ↵
      1. Wang Y-F,
      2. Zhang Y,
      3. Lin Z, et al
      . Identification of 38 novel loci for systemic lupus erythematosus and genetic heterogeneity between ancestral groups. Nat Commun 2021;12:772. doi:10.1038/s41467-021-21049-y
    21. ↵
      1. Hochberg MC
      . Updating the American college of rheumatology revised criteria for the classification of systemic lupus erythematosus. Arthritis Rheum 1997;40:1725. doi:10.1002/art.1780400928
    22. ↵
      1. Petri M,
      2. Orbai A-M,
      3. Alarcón GS, et al
      . Derivation and validation of the systemic lupus international collaborating clinics classification criteria for systemic lupus erythematosus. Arthritis Rheum 2012;64:2677–86. doi:10.1002/art.34473
      OpenUrlCrossRefPubMedWeb of Science
    23. ↵
      1. Fan CT,
      2. Lin JC,
      3. Lee CH
      . Taiwan Biobank: a project aiming to aid Taiwan’s transition into a biomedical island. Pharmacogenomics 2008;9:235–46. doi:10.2217/14622416.9.2.235
      OpenUrlCrossRefPubMedWeb of Science
    24. ↵
      1. Choi SW,
      2. Mak TS-H,
      3. O’Reilly PF
      . Tutorial: a guide to performing polygenic risk score analyses. Nat Protoc 2020;15:2759–72. doi:10.1038/s41596-020-0353-1
      OpenUrlPubMed
    25. ↵
      1. Wray NR,
      2. Goddard ME,
      3. Visscher PM
      . Prediction of individual genetic risk to disease from genome-wide association studies. Genome Res 2007;17:1520–8. doi:10.1101/gr.6665407
      OpenUrlAbstract/FREE Full Text
    26. ↵
      1. Choi SW,
      2. O’Reilly PF
      . PRSice-2: polygenic risk score software for biobank-scale data. Gigascience 2019;8:giz082. doi:10.1093/gigascience/giz082
    27. ↵
      1. Privé F,
      2. Arbel J,
      3. Vilhjálmsson BJ, et al
      . LDpred2: better, faster, stronger. Bioinformatics 2021;36:5424–31. doi:10.1093/bioinformatics/btaa1029
      OpenUrl
    28. ↵
      1. Mak TSH,
      2. Porsch RM,
      3. Choi SW, et al
      . Polygenic scores via penalized regression on summary statistics. Genet Epidemiol 2017;41:469–80. doi:10.1002/gepi.22050
      OpenUrl
    29. ↵
      1. Ge T,
      2. Chen C-Y,
      3. Ni Y, et al
      . Polygenic prediction via Bayesian regression and continuous shrinkage priors. Nat Commun 2019;10:1776. doi:10.1038/s41467-019-09718-5
    30. ↵
      1. Gladman DD,
      2. Ibañez D,
      3. Urowitz MB
      . Systemic lupus erythematosus disease activity index 2000. J Rheumatol 2002;29:288–91.
      OpenUrlAbstract/FREE Full Text
    31. ↵
      1. Bajema IM,
      2. Wilhelmus S,
      3. Alpers CE, et al
      . Revision of the International society of nephrology/renal pathology society classification for lupus nephritis: clarification of definitions, and modified National Institutes of Health activity and chronicity indices. Kidney Int 2018;93:789–96. doi:10.1016/j.kint.2017.11.023
      OpenUrlCrossRefPubMed
    32. ↵
      1. Boddaert J,
      2. Huong DLT,
      3. Amoura Z, et al
      . Late-onset systemic lupus erythematosus: a personal series of 47 patients and pooled analysis of 714 cases in the literature. Medicine (Baltimore) 2004;83:348–59. doi:10.1097/01.md.0000147737.57861.7c
      OpenUrlCrossRefPubMedWeb of Science
    33. ↵
      1. Yin X,
      2. Kim K,
      3. Suetsugu H, et al
      . Meta-analysis of 208370 East Asians identifies 113 susceptibility loci for systemic lupus erythematosus. Ann Rheum Dis 2021;80:632–40. doi:10.1136/annrheumdis-2020-219209
      OpenUrlAbstract/FREE Full Text
    34. ↵
      1. Khunsriraksakul C,
      2. Li Q,
      3. Markus H, et al
      . Multi-ancestry and multi-trait genome-wide association meta-analyses inform clinical risk prediction for systemic lupus erythematosus. Nat Commun 2023;14:668. doi:10.1038/s41467-023-36306-5
    35. ↵
      1. Thareja G,
      2. Belkadi A,
      3. Arnold M, et al
      . Differences and commonalities in the genetic architecture of protein quantitative trait loci in European and Arab populations. Hum Mol Genet 2023;32:907–16. doi:10.1093/hmg/ddac243
      OpenUrl
    36. ↵
      1. Chen L,
      2. Wang Y-F,
      3. Liu L, et al
      . Genome-wide assessment of genetic risk for systemic lupus erythematosus and disease severity. Hum Mol Genet 2020;29:1745–56. doi:10.1093/hmg/ddaa030
      OpenUrlPubMed
    37. ↵
      1. Lisnevskaia L,
      2. Murphy G,
      3. Isenberg D
      . Systemic lupus erythematosus. Lancet 2014;384:1878–88. doi:10.1016/S0140-6736(14)60128-8
      OpenUrlCrossRefPubMedWeb of Science
    38. ↵
      1. Tomic-Lucic A,
      2. Petrovic R,
      3. Radak-Perovic M, et al
      . Late-onset systemic lupus erythematosus: clinical features, course, and prognosis. Clin Rheumatol 2013;32:1053–8. doi:10.1007/s10067-013-2238-y
      OpenUrlCrossRefPubMed
    39. ↵
      1. Sousa S,
      2. Gonçalves MJ,
      3. Inês LS, et al
      . Clinical features and long-term outcomes of systemic lupus erythematosus: comparative data of childhood, adult and late-onset disease in a national register. Rheumatol Int 2016;36:955–60. doi:10.1007/s00296-016-3450-2
      OpenUrl
    40. ↵
      1. Taylor KE,
      2. Remmers EF,
      3. Lee AT, et al
      . Specificity of the STAT4 genetic association for severe disease manifestations of systemic lupus erythematosus. PLoS Genet 2008;4:e1000084. doi:10.1371/journal.pgen.1000084
    41. ↵
      1. Taylor KE,
      2. Chung SA,
      3. Graham RR, et al
      . Risk alleles for systemic lupus erythematosus in a large case-control collection and associations with clinical subphenotypes. PLoS Genet 2011;7:e1001311. doi:10.1371/journal.pgen.1001311
    42. ↵
      1. Gladman D,
      2. Ginzler E,
      3. Goldsmith C, et al
      . The development and initial validation of the systemic lupus international collaborating clinics/American college of rheumatology damage index for systemic lupus erythematosus. Arthritis Rheum 1996;39:363–9. doi:10.1002/art.1780390303
      OpenUrlCrossRefPubMedWeb of Science

    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.

    Footnotes

    • Y-MC and H-IY contributed equally.

    • Contributors H-IY, Y-MC and Y-JC conceived and designed the study. H-IY, Y-MC, Y-JC, Y-CL, W-JJ and Y-CH performed the literature search and interpretation of data. T-HH, C-LM, C-YW and I-CC conducted data extraction, methodological quality assessments and performed the analysis. Y-JC drafted the manuscript. H-IY and Y-MC performed critical revision of the manuscript for important intellectual content. H-IY is the guarantor, responsible for the overall content. All authors read and approved the final version of submitted manuscript.

    • Funding This study was supported by a grant from Taichung Veterans General Hospital, Taiwan (TCVGH-1113801B).

    • Competing interests None declared.

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

    • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.