Design: retrospective multicentre registry analysis with multivariable regression, sensitivity analyses and propensity score matching
Population: 1,423 newly initiated biological or targeted synthetic DMARD courses continuing ≥12 months with rheumatoid factor measured at baseline and three months
Primary outcome: early rheumatoid factor change by drug class, and association with twelve-month Clinical Disease Activity Index remission
Effect: Janus kinase inhibitors vs tumour necrosis factor inhibitors β = −14.50 (95% CI −20.04 to −8.97); ≥20% early decline associated with 12-month remission, adjusted OR 1.39 (1.04–1.88); trend across reduction categories P = 0.004

Rheumatoid factor is normally treated as a fixed diagnostic and prognostic attribute measured once. The ANSWER cohort asked whether it moves with treatment and whether the movement means anything, analysing 1,423 newly started biological or targeted synthetic disease-modifying drug courses that continued for at least twelve months and had rheumatoid factor measured at baseline and three months.

It moves, and it differs by drug class. Janus kinase inhibitors produced the largest early decline; against tumour necrosis factor inhibitors the difference was β = −14.50 (95% confidence interval −20.04 to −8.97, P < 0.001), a direction that held in sensitivity analyses and in propensity score-matched analysis. An early decline — defined as at least a 20% fall at three months — was independently associated with twelve-month Clinical Disease Activity Index remission (adjusted odds ratio 1.39, 95% confidence interval 1.04 to 1.88, P = 0.029), and remission rates rose across ordered categories of greater rheumatoid factor reduction (P for trend = 0.004).

The odds ratio is modest, and this is a prognostic marker rather than a treatment target. What it offers is a cheap extra piece of information at a three-month review that most units are already taking blood at. It does not replace the disease activity score, which remains the thing to act on — but where the clinical picture at three months is ambiguous, a rheumatoid factor that has not moved at all is a piece of evidence pointing one way.

Repeat rheumatoid factor at three months when starting a biological or targeted synthetic drug, alongside the usual monitoring bloods.
Use a 20% fall as the threshold; that is what the association was defined on.
Treat it as supporting information, not as a reason to switch — the disease activity score remains the decision variable.
Expect larger falls on Janus kinase inhibitors, so compare against the drug class rather than in the abstract.
Do not repeat it in a seronegative patient; there is nothing to follow.

Why it matters

A test we order once and treat as fixed turns out to move with treatment and to carry prognostic information when it does.

The statistics, in plain English

An adjusted odds ratio of 1.39 with a confidence interval from 1.04 to 1.88 only just excludes 1.0, so this is a weak association and should be used as supporting information rather than as a decision rule. The trend across ordered categories (P = 0.004) is more persuasive than the binary threshold, because a dose-response pattern is harder to produce by chance. The class comparison is retrospective: patients started on a Janus kinase inhibitor differ from those started on a tumour necrosis factor inhibitor in ways propensity matching narrows but does not remove, and the cohort required twelve months of continued treatment — so everyone who stopped early is absent from the analysis.