Rheumatoid arthritis — statistics sheet

Last curated: 2026-11-01. Figures are not pooled across incompatible definitions. Every row states source, year, population and method.

Population burden

Metric	Figure	Population/year	Method	Source
Global prevalent cases	17.6 million	204 countries/territories, 2020	GBD 2021 modelling	GBD 2021 RA Collaborators 2023, PMID 37795020
Age-standardized prevalence	208.8 per 100,000	Global, 2020	GBD reference standard	PMID 37795020
Projected prevalent cases	31.7 million	Global, 2050	Demographic/GBD projection	PMID 37795020
Absolute projected increase	14.1 million	2020→2050	Difference of reported counts	PMID 37795020
Relative projected increase	~80%	2020→2050	31.7/17.6 − 1	PMID 37795020
Countries in earlier GBD series	195	1990–2017	GBD 2017	Safiri 2019, PMID 31511227
LMIC prevalent cases	12.6 million	129 low- and middle-income countries, 2023	GBD 2023 modelling	Wang 2026, PMID 42495527
Upper-middle- versus low-income prevalence ratio	4.1-fold	129 LMICs, 2023	GBD 2023 modelling stratified by gross national income	PMID 42495527
Adult worldwide prevalence (review)	0.53% of adults	Worldwide adults	JAMA narrative review, 2026	Smolen 2026, PMID 42720931
US adult prevalence (review)	0.74% of adults	US adults	Same review	PMID 42720931

Classification and measurement thresholds

Metric	Figure	Population/year	Method	Source
2010 classification threshold	≥6/10	Eligible synovitis population	ACR/EULAR criteria development	Aletaha 2010, PMID 20872595
Joint-domain maximum	5 points	>10 joints including ≥1 small joint	Criteria score	PMID 20872595
Serology maximum	3 points	High-positive RF or ACPA	Criteria score	PMID 20872595
Duration contribution	1 point	Symptoms ≥6 weeks	Criteria score	PMID 20872595
Validation cohort	270	Recent-onset arthritis, 2-year follow-up	Prospective cohort	Varache 2011, PMID 21572146
Boolean 2.0 tender joints	≤1	Trial validation datasets	Remission definition	Studenic 2023, PMID 36274193
Boolean 2.0 swollen joints	≤1	Trial validation datasets	Remission definition	PMID 36274193
Boolean 2.0 CRP	≤1 mg/dL	Trial validation datasets	Remission definition	PMID 36274193
Boolean 2.0 patient global	≤2/10	Trial validation datasets	Revised threshold	PMID 36274193
SDAI remission	≤3.3	RA activity assessment	Index threshold	PMID 36274193
CDAI remission	≤2.8	RA activity assessment	Index threshold	Messelink 2023, PMID 37116986
DAS28 remission	<2.6	RA activity assessment	Historical index threshold	Messelink 2023, PMID 37116986

Strategy and treatment-trial scale

Study/metric	Figure	Population/year	Method	Source
TICORA randomized	111	Active RA, 2004	Single-blind RCT	Grigor 2004, PMID 15262104
TICORA screened	183	Two teaching hospitals	Trial screening	PMID 15262104
FIN-RACo randomized cohort	195	Early active RA	Combination vs single DMARD RCT/follow-up	Korpela 2004, PMID 15248204
ATTRACT randomized	428	Active despite MTX	54-week RCT	Lipsky 2000, PMID 11096166
PREMIER randomized	799	Early aggressive MTX-naive RA	2-year RCT	Breedveld 2006, PMID 16385520
TEMPO randomized	686	Active RA	Double-blind RCT	Klareskog 2004, PMID 15001324
AMBITION randomized	673	Active RA	24-week RCT	Jones 2010, PMID 19297346
SELECT-COMPARE randomized	1,629	MTX inadequate response	Phase 3 RCT	Fleischmann 2019, PMID 31287230
R4RA randomized	164	Anti-TNF inadequate response	Biopsy-stratified phase 4 RCT	Humby 2021, PMID 33485455
Biologic poor response	~40%	Established RA	R4RA background estimate	PMID 33485455
Failure after ≥4 advanced therapies	1.6%	EULAR-defined D2T cohorts	Meta-analysis	Xie 2026, PMID 41188120

Safety and systemic burden

Metric	Figure	Population/year	Method	Source
ORAL Surveillance randomized	4,362	RA age ≥50 plus ≥1 CV risk factor	Active-comparator safety RCT	Ytterberg 2022, PMID 35081280
Tofacitinib doses tested	5 or 10 mg twice daily	Same trial	Randomized dose arms	PMID 35081280
Mortality meta-analysis studies	17	RA cohorts	Meta-analysis	Lee 2024, PMID 38918258
Mortality meta-analysis patients	486,098	Multiple countries/eras	Meta-analysis	PMID 38918258
Deaths included	63,988	Same	Meta-analysis	PMID 38918258
RA-ILD share of RA mortality	10–20%	Review-level estimate	Literature synthesis	Cassone 2020, PMID 32290218
RA-ILD mean survival cited in review	5–8 years	Historical/heterogeneous RA-ILD cohorts	Review synthesis	PMID 32290218
RA-ILD JAK observational studies	7	Published through review search	Systematic review	Narváez 2024, PMID 39270812
RA-ILD incident-risk studies in same review	3	RA cohorts	Systematic review	PMID 39270812
RA-ILD prediction derivation cohort	1,156	Single center	Retrospective model	Yao 2025, PMID 41299487
Derivation RA-ILD cases	400	Same	Retrospective model	PMID 41299487
External validation cases/controls	178/178	Chinese registry	External validation	PMID 41299487

Difficult-to-treat, fatigue and precision

Metric	Figure	Population/year	Method	Source
Reported D2T prevalence range	5.5–27.5%	Heterogeneous RA cohorts	Review synthesis	Hofman 2025, PMID 39383505
D2T poly-refractory cohort	1,591	b/tsDMARD-treated RA	Cross-sectional ultrasound study	David 2024, PMID 38059326
Blinatumomab compassionate series	6	Multidrug-resistant RA	Case series	Bucci 2024, PMID 38671240
Severe fatigue	at least 1 in 6	Treated RA literature	Review synthesis	Pope 2020, PMID 32385141
Residual-symptom review reports	55	Literature through 2018	Systematic review	Michaud 2021, PMID 32619340
Unique studies in residual review	53	Same	Systematic review	PMID 32619340
Plasma-proteome RA cohort	278	Longitudinal RA	Proteomics cohort	He 2025, PMID 40691443
At-risk participants in proteome study	60	Same	Longitudinal cohort	PMID 40691443
Healthy controls in proteome study	99	Same	Cohort	PMID 40691443
Baricitinib 24-month retention	58.2% overall; 42.0% D2T	Japan; 353 total, 88 D2T	24-month observational cohort	Ikeda 2026, PMID 42667647
Baricitinib CDAI remission	36.4% overall; 22.9% D2T	Same	Observational outcome	PMID 42667647
TNFA methylation model AUC	0.760 combined versus 0.699 clinical-only	201 early untreated RA	Treatment-response biomarker cohort	Dalix 2026, PMID 42664067

Periodic-sweep additions — treatment and systemic outcomes

Metric	Figure	Population/year	Method	Source
Glucocorticoid use at 12 months	47% oral; 26% parenteral; 8% no early glucocorticoid	2,222 Canadian early-RA participants, 2007–2023	Observational cohort; exposure in first 3 months	Fernández-Codina 2026, PMID 42665531
Adjusted odds of glucocorticoid use at 12 months	OR 9.8 oral; 4.1 parenteral, versus no early glucocorticoid	Same	Multivariable logistic regression	PMID 42665531
CKD-stage progression	106/601 over five years	Japanese RA cohort	Retrospective cohort	Sugiyama 2026, PMID 42667508

Prevention trials

Metric	Figure	Population/year	Method	Source
ARIAA treatment duration	6 months	ACPA-positive arthralgia + MRI inflammation	Placebo-controlled RCT	Rech 2024, PMID 38364841
ARIAA centers	14	Germany, Spain, Czech Republic	Multicenter RCT	PMID 38364841
APIPPRA UK centers	28	Autoantibody-positive inflammatory joint pain	Phase 2b RCT	Cope 2024, PMID 38364839
APIPPRA Netherlands centers	3	Same	Phase 2b RCT	PMID 38364839
TREAT EARLIER treatment duration	1 year	Arthralgia + MRI subclinical inflammation	Proof-of-concept RCT	Krijbolder 2022, PMID 35871815
TREAT EARLIER follow-up	4 years	ACPA-negative subgroup analysis	Long-term follow-up	Dumoulin 2024, PMID 39303731
TREAT EARLIER five-year overall RA onset	22% treatment vs 27% placebo; HR 0.79 (95% CI 0.47–1.34)	236 clinically suspect arthralgia participants	Randomized follow-up	Mulligen 2026, PMID 42392130
TREAT EARLIER increased-risk ACPA-negative subgroup	9% treatment vs 32% placebo; HR 0.24 (95% CI 0.07–0.87); NNT 4	66 participants; 5 years	Prespecified risk-stratified follow-up	PMID 42392130
APIPPRA/ALTO arthritis-free-survival difference	4.9 months (95% CI 0.1–9.6; p=0.044) at 4 years	143 long-term participants; median follow-up 55 months	Masked randomized follow-up	Cope 2026, PMID 41576971
StopRA clinical RA at 36 months	30.4% HCQ vs 32.9% placebo; risk difference −0.058 (95% CI −0.336 to 0.220)	142 modified-ITT anti-CCP3-high participants	Phase 2 RCT	Deane 2026, PMID 40884017

Known conflicts and caveats

• GBD figures are modelled and should not be averaged with claims or examination-survey prevalence.
• RA-ILD pooled prevalence is shown with I² and study design because systematic HRCT, clinical coding and symptomatic cohorts estimate different constructs; 18.7% is not a universal clinical prevalence.
• Historical cervical-spine prevalence up to 80% comes from severe-era cohorts and is not a modern population estimate (Joaquim 2014, PMID 25151973).
• D2T prevalence depends on whether symptoms, objective inflammation and access-related failure are included.
• Trial enrollment is not an effect size; it is shown to indicate evidence scale.
• Drug-class event rates cannot be transported across baseline cardiovascular, infection or cancer risk without absolute-risk recalculation.

Deepening additions — population and diagnostic performance

Metric	Figure	Population/year	Method	Source
Global pooled prevalence	0.46% (95% CI 0.39–0.54)	67 studies; 742,246 cases/211.6 million controls; literature 1980–2019	Random-effects meta-analysis; I²=99.9%, prediction interval 0.06–1.27%	Almutairi 2021, PMID 33175207
Linked-data prevalence subgroup	0.69% (95% CI 0.47–0.95)	Record-linkage studies	Meta-analysis subgroup	PMID 33175207
2010 criteria sensitivity/specificity	0.82 (95% CI 0.79–0.84) / 0.61 (0.59–0.64)	6,816 patients in 17 full articles	Systematic review/meta-analysis	Radner 2014, PMID 23592710
ACPA sensitivity/specificity	67% (95% CI 62–72) / 95% (94–97)	37 studies	Diagnostic meta-analysis	Nishimura 2007, PMID 17548411
ACPA positive/negative likelihood ratios	12.46 / 0.36	Same	Diagnostic meta-analysis	PMID 17548411
RF sensitivity/specificity	69% (95% CI 65–73) / 85% (82–88)	50 studies	Diagnostic meta-analysis	PMID 17548411
Major depression prevalence	16.8% (95% CI 10–24)	72 studies; 13,189 patients	Diagnostic-definition meta-analysis	Matcham 2013, PMID 24003249
PHQ-9-defined depression	38.8% (95% CI 34–43)	RA cross-sectional studies	Screening-instrument subgroup	PMID 24003249

Deepening additions — treatment and strategy effects

Study/metric	Figure	Population/year	Method	Source
RACAT DAS28 change	−2.1 triple therapy vs −2.3 etanercept–MTX; upper 95% confidence limit 0.41 below 0.6 margin	353 MTX inadequate responders; 48 weeks	Double-blind noninferiority RCT	O'Dell 2013, PMID 23755969
TEAR week-102 radiographic change	0.64 etanercept–MTX vs 1.69 triple therapy; p=0.047	Early aggressive RA; 2 years	Randomized strategy trial	Moreland 2012, PMID 22508468
CAMERA ever-remission	50% intensive vs 37% conventional; p=0.03	299 early RA; 2 years	Open-label randomized strategy trial	Verstappen 2007, PMID 17519278
BeSt one-year radiographic progression	Median 2.0, 2.5, 1.0 and 0.5 across four strategies; p<0.001	508 early RA	Randomized strategy trial	Goekoop-Ruiterman 2005, PMID 16258899
NORD-STAR CDAI remission	39.2% conventional; 59.3% abatacept; 52.3% certolizumab; 51.9% tocilizumab	812 treatment-naive early RA; week 48	Randomized active-comparator trial	Østergaard 2023, PMID 37423647
FINCH 2 ACR20	66.0% filgotinib 200 mg, 57.5% 100 mg, 31.1% placebo	448 biologic inadequate responders; week 12	Phase 3 placebo-controlled RCT	Genovese 2019, PMID 31334793
ARCTIC primary endpoint	22% ultrasound target vs 19% clinical target; difference 3.3% (95% CI −7.1 to 13.7)	230 early RA; 2 years	Randomized strategy trial	Haavardsholm 2016, PMID 27530741
TaSER DAS44 change	−2.69 ultrasound vs −2.58 clinical; between-group 95% CI −0.70 to 0.48	111 early RA; 18 months	Randomized strategy trial	Dale 2016, PMID 27026689

Deepening additions — systemic disease and safety

Metric	Figure	Population/year	Method	Source
MI incidence-rate ratio	2.10 (95% CI 1.52–2.89)	17 cohorts; 124,894 RA patients	Meta-analysis	Meune 2010, PMID 20656636
Stroke incidence-rate ratio	1.91 (95% CI 1.73–2.12)	Same	Meta-analysis	PMID 20656636
RA-ILD pooled prevalence	18.7% (95% CI 15.8–21.6); I²=96.4%	56 studies; 11,851 RA-ILD cases	Meta-analysis of heterogeneous definitions	Wang 2024, PMID 38547537
RA-ILD UIP mortality HR	1.88 (95% CI 1.14–3.10)	23 eligible studies	Prognostic meta-analysis	Qiu 2021, PMID 34635095
INBUILD RA-ILD FVC decline	−82.6 vs −199.3 mL/year; difference 116.7 (95% CI 7.4–226.1)	89 progressive RA-ILD participants; 52 weeks	Randomized trial subgroup	Matteson 2023, PMID 37209188
RA background VTE OR	2.23 overall; DVT 2.25; PE 2.15	272,884 RA/2.28 million controls	Meta-analysis	Hu 2021, PMID 34859863
Influenza satisfactory vaccine response	75.5% holding MTX 2 weeks vs 54.5% continuing; p<0.001	316 RA patients	Randomized trial	Park 2018, PMID 29572291
RA pregnancy preterm-delivery OR	1.83	11,999 RA pregnancies/9.92 million controls	Meta-analysis	Sim 2023, PMID 36544253
D2T pooled prevalence	11.7% (95% CI 9.5–14.3)	23 studies; 27,987 patients; 13 countries	EULAR-definition meta-analysis	Xie 2026, PMID 41188120
Persistently inflammatory share of D2T	47.1% (95% CI 33.3–61.4)	D2T cohorts	Meta-analysis	PMID 41188120

Known conflicts and caveats — deepening addendum

• Diagnostic sensitivity and specificity depend on the eligible population and reference standard; classification criteria, ACPA and RF rows are not screening performance in unselected pain.
• The global 0.46% pooled prevalence has I²=99.9% and a 0.06–1.27% prediction interval; it must not replace a locally validated estimate.
• Strategy-trial effects combine drug sequence, visit frequency, escalation rules and era; they are not pure drug effects.
• RA-ILD estimates deliberately retain design and heterogeneity. HRCT-detected abnormalities, clinically coded ILD and progressive fibrotic RA-ILD are different denominators.
• Relative safety estimates require baseline absolute risk, comparator, geography, dose and exposure time; “not statistically significant” does not establish equivalence.
• JAMA adult prevalence (0.53% worldwide / 0.74% US) uses an adult denominator and is not interchangeable with GBD all-age modelled prevalence.

2026-10-11 sweep additions

Metric	Figure	Population/year	Method	Source
b/tsDMARD taper vs continuation, LDA at 24 weeks	88.2% vs 89.6%; RD −1.43% (95% CI −8.09 to 5.23)	348 people with DAS28-ESR LDA ≥6 months	Open-label RCT, non-inferiority margin −10%	Park 2026, PMID 42839723
Any DMARD tapering, flare RR	1.56 (95% CI 1.26–1.98)	27 publications; 5,262 participants	Systematic review/meta-analysis of RCTs	Zhang 2026, PMID 42836514
Any DMARD withdrawal, flare RR	2.23 (95% CI 1.83–2.94)	Same	Same	PMID 42836514
Stop bDMARD / continue csDMARD, flare RR vs maintenance	3.06 (95% CI 1.81–5.16)	19 studies; 3,115 people in sustained remission	Frequentist network meta-analysis	Tandi 2026, PMID 42843067
Reduce bDMARD / continue csDMARD, flare RR vs maintenance	1.17 (95% CI 0.76–1.82)	Same	Same	PMID 42843067
First-line b/tsDMARD+MTX vs MTX, DAS28 remission at 6 months	RR 1.62 (95% CI 1.39–1.89)	29 trials; 13,387 MTX-naive early RA	Pairwise meta-analysis	Russell 2026, PMID 42755040
Same comparison at 12 months	RR 1.54 (95% CI 1.39–1.71)	Same	Same	PMID 42755040
Residual swollen joints in RA remission	5–25%	Residual-disease SR; 84 prevalence studies	Systematic review/meta-analysis	Webers 2026, PMID 42485861
JAK ACR20 RR, class 3 obesity vs healthy weight	0.78 (95% CI 0.71–0.85)	16 phase 3 trials; 11,883 people	IPD meta-analysis	Bechman 2026, PMID 42546722
JAK vs TNF, MACE in age ≥65	HR 1.13 (95% CI 0.91–1.41)	1,304 matched pairs; up to 3 years	Propensity-matched cohort	Ibis 2026, PMID 42823079
JAK vs TNF, VTE in age ≥65	HR 1.18 (95% CI 0.79–1.78)	Same	Same	PMID 42823079
Blinatumomab week-12 CDAI LDA / remission	9/15 LDA; 3/15 remission	15 named-patient MDR-RA	Case series	Bucci 2026, PMID 42595657
Blinatumomab subsequent flare	14/15	Same	Same	PMID 42595657
Tofacitinib RA-ILD any AE / deaths / PPF-or-death	36/39; 3 deaths; 15/39	39 people; 48 weeks	Single-arm phase II (NCT05246293)	Rojas-Serrano 2026, PMID 42722091
Tofacitinib RA-ILD FVC change	−72 mL at 48 weeks	Same	Same	PMID 42722091
APIPPRA external validation of 14 risk rules	10/14 C-statistic <0.6; 4/14 0.6–0.7	APIPPRA interception cohort	External validation	Jasenecova 2026, PMID 42676706
ATI-450 week-12 ACR20	42.2% (20 mg); 34.7% (50 mg); 46.5% placebo	251 people; phase 2	ClinicalTrials.gov posted results, 2026-09-28	NCT05279417

2026-10-18 sweep additions

Metric	Figure	Population/year	Method	Source
ARCTIC 10-year DAS remission	82%	179 of 218 surviving ARCTIC participants (230 enrolled), 10 years after early-RA diagnosis	Long-term follow-up of a 2-year treat-to-target RCT cohort	Nordberg 2026, PMID 42881069
ARCTIC 10-year radiographic progression	Median vdH-mSharp change 2.0 (IQR 0.5–4.5)	Same	Same	PMID 42881069
JAK vs TNF, nonfatal MI or ischemic stroke (cardiovascular death excluded), ORAL Surveillance-like population	HR 1.002 (95% CI 0.923–1.088); +2.11 per 1,000 at 5 years	18,757 matched new-user pairs, age ≥50 with ≥1 CV risk factor	TriNetX target-trial emulation (observational; PubMed mis-indexes as RCT)	McClenahan 2026, PMID 42877103
JAK vs TNF, all-cause mortality (secondary)	HR 1.158 (95% CI 1.058–1.268); +10.4 per 1,000; E-value 1.586	Same	Same	PMID 42877103
CD19 CAR-T (miv-cel) safety	CRS 6/6 (grade 1–2); ICANS 0/6; 1 dose-limiting grade 3 transaminase rise	6 people with refractory ACPA-positive RA; 36–52 weeks	Nonrandomized phase 1 (NCT06475495)	Albach 2026, PMID 42661098
CD19 CAR-T efficacy	Median DAS28-CRP −34%; DAS28-CRP remission and ACR70 3/6; ACPA (anti-MCV) seroconversion 4/6; IgM-RF 5/6	Same	Same	PMID 42661098
JAK cycling vs biologic switch after first JAK failure, 12-month discontinuation	HR 0.27 (95% CI 0.10–0.71)	186 people (44 cycled, 142 switched)	Multicenter retrospective cohort, IPTW	Tsai 2026, PMID 42869207
Lung ultrasound for RA-ILD (≥5 B-lines)	Sensitivity 60.9%, specificity 63.7%, NPV 85%; negative LUS + normal DLCO NPV 100% (95% CI 82–100)	103 at-risk RA patients; HRCT ILD in 22.3%	Cross-sectional diagnostic accuracy vs HRCT	Kouna 2026, PMID 42873714
RF seroprevalence	9.3%; stable in 94.5%	469,036 UK Biobank participants with RF testing	Population biobank serology	Hocaoğlu 2026, PMID 42378266
CHIP and incident seropositive RA	HR 1.26 (95% CI 1.03–1.52); DNMT3A-CHIP late-onset HR 1.45 (1.13–1.96); no association with seronegative RA	612,989 participants in All of Us, BioVU and UK Biobank	Pooled biobank survival analysis	Zhao 2026, PMID 41847932
Early inflammatory arthritis seen ≤3 weeks / DMARD ≤6 weeks of referral	42% / 53%	10,695 patients from 122 of 127 eligible providers, England, Wales and Jersey, April 2025–March 2026 (all newly diagnosed EIA, not RA alone)	National clinical audit (NEIAA Year 8)	(British Society for Rheumatology / NEIAA — "State of the Nation Summary Report 2026", https://www.rheumatology.org.uk/Portals/0/Documents/Practice_Quality/Audit/NEIA/2026/National_Early_Inflammatory_Autoimmune_Diseases_Audit_State_of_the_Nation_Summary_Report_2026.pdf, accessed 2026-10-18)
EIA DAS28 remission at 3 months	33% (50% of eligible records had a 3-month DAS28)	Same	Same	Same
Same-day DMARD start and 3-month remission	38% vs 27%; adjusted odds 48% higher	Same	Same	Same

2026-10-25 sweep additions

Metric	Figure	Population/year	Method	Source
RA-ILD incidence, Optum EHR	3.9 per 100,000 person-years (95% CI 3.7–4.1)	US eligible database population, 2019–2024; denominator is the full base population, not only people with RA	Diagnosis-code identification; age/sex adjustment to US Census population; full-text Table 2, RA-ILD column	Martins 2026, PMID 42909770
RA-ILD incidence, MarketScan	5.3 per 100,000 person-years (95% CI 4.8–5.9)	US Commercial/Medicare database base population, 2019–2024; not an RA-only denominator	Same method; full-text Table 2, separate database estimate	Martins 2026, PMID 42909770

The two database estimates are retained separately. They measure coded RA-ILD incidence in the respective base populations, not the proportion of screened RA patients with HRCT abnormalities. The abstract's overall CTD-ILD rates must not be substituted for these RA-specific values. The earlier CARDINAL-RA row is clarified to name its nonfatal-event composite after full-text verification (McClenahan 2026, PMID 42877103).

2026-11-01 sweep additions

Metric	Figure	Population/year	Method	Source
bDMARD discontinuation vs dose reduction/maintenance, flare RR	2.13 (95% CI 1.74–2.61)	11 studies (10 RCTs + 1 extension); 2,861 people; search to 1 Feb 2026	Pairwise random-effects meta-analysis	Lee 2026, PMID 42269230
Same, early RA (<2 years)	RR 1.62 (95% CI 1.25–2.10)	Subgroup of the same synthesis	Same	PMID 42269230
Same, established RA	RR 2.31 (95% CI 1.89–2.82)	Same	Same	PMID 42269230
Recapture after bDMARD restart	87.5% (95% CI 82.4–91.6); median 12.4 weeks	Same	Pooled capture rate	PMID 42269230
Anti-TNF tapering, remission RR / flare RR	0.69 (0.57–0.84) / 1.96 (1.57–2.45)	15 RCTs; 2,782 people; search Feb 2024	Pairwise meta-analysis of taper vs standard dose	Rotea-Salvo 2026, PMID 42963167
JAK-inhibitor tapering, LDA RR / remission RR	0.83 (0.76–0.91) / 0.86 (0.75–0.99)	Same review	Same	PMID 42963167
JAK1 vs pan-JAK discontinuation HR	0.72 (95% CI 0.61–0.85)	7 observational studies; 21,244 people	Real-world retention meta-analysis	Lee 2026, PMID 42034889
JAK1 vs pan-JAK, lack-of-efficacy discontinuation HR	0.68 (95% CI 0.55–0.84)	Same	Same	PMID 42034889
TNF vs JAK serious infection HR	0.55 (95% CI 0.39–0.77); E-value 2.39	5,018 Japanese claims new users (4,042 TNF / 976 JAK)	Overlap-weighted cohort	Ota 2026, PMID 41422438
NZ GP age-standardised RA incidence	26.9 per 100,000 (95% CI 22.8–30.9)	Pinnacle general-practice dataset, 2018–2025; 3,831 people with RA	WHO age-standardised SNOMED-coded incidence	Lao 2026, PMID 42490391
NZ GP age-standardised RA prevalence	417.5 per 100,000 (95% CI 402.8–432.1)	Same	Same, prevalence	PMID 42490391
NZ incident RA SMR vs general population	1.07 (95% CI 0.80–1.34)	Incident RA in the same dataset	Standardised mortality ratio	PMID 42490391