Design: multicentre registry analysis using group-based trajectory modelling with landmark logistic regression at weeks 4 and 12, multiple imputation and bootstrap internal validation
Population: 843 patients with rheumatoid arthritis treated with Janus kinase inhibitors, followed over 52 weeks
Primary outcome: prediction of a favourable 52-week disease activity trajectory from early Clinical Disease Activity Index values
Effect: four trajectories — rapid responders 64.9%, partial responders 18.1%, high-baseline responders 12.7%, persistent non-responders 4.3%; week-52 remission 0.0% to 47.7%; optimal thresholds CDAI 12.0 at week 4 and 9.3 at week 12, with the absolute score outperforming the slope

The twelve-week review on a Janus kinase inhibitor is where the continue-or-switch decision is usually made, and it is usually made on an impression of how much things have improved. A registry analysis of 843 patients with rheumatoid arthritis asked what actually predicts the year ahead, using group-based trajectory modelling of the Clinical Disease Activity Index over 52 weeks and landmark analyses at weeks 4 and 12.

Four trajectories emerged: rapid responders (64.9%), partial responders with persistent residual activity (18.1%), responders starting from high baseline activity (12.7%) and persistent non-responders (4.3%). Week-52 remission rates across those groups ran from 0.0% to 47.7% — the non-responder group reached remission in nobody. Adding the disease activity index at the landmark time points substantially improved prediction of a favourable trajectory, and the optimal thresholds were 12.0 at week 4 and 9.3 at week 12. Bootstrap internal validation showed minimal optimism and good calibration.

The finding that changes practice is which metric carried the information. The absolute score at the landmark outperformed the slope; adding the rate of change contributed almost nothing once the week-12 score was known. That is the opposite of how the review conversation usually goes, which is dominated by how much better the patient is than at baseline. A patient who has improved a great deal but sits above a Clinical Disease Activity Index of 9.3 at twelve weeks is in a worse position than one who has improved less and sits below it. Two cautions: these thresholds come from one registry with internal validation only, and patients starting from very high activity formed a distinct trajectory group, so a single cut-off should inform the conversation rather than close it.

Record the Clinical Disease Activity Index at weeks 4 and 12 as fixed review points after starting a Janus kinase inhibitor.
Judge the week-12 review on where the score sits, not on how far it has fallen.
Treat a week-12 score above about 9.3 as a reason to plan the next step rather than wait.
Allow for the high-baseline trajectory group, who improved but along a different path.
Remember the thresholds are internally validated in one registry — use them to inform the decision, not to make it.

Why it matters

The twelve-week review is conducted around how much better the patient is, and the measurement that predicts the year is where they have actually got to.

The statistics, in plain English

Bootstrap internal validation with minimal optimism means the model is not badly overfitted to its own data, which is reassuring but is not the same as external validation — thresholds derived in one registry typically shift when applied elsewhere. The spread in week-52 remission across trajectory groups (0.0% to 47.7%) shows the trajectories are clinically meaningful and not a statistical artefact. The most important methodological point is the comparison between absolute score and slope: Brier scores improved markedly when the week-12 score was added and barely at all when the slope was added afterwards, which is a direct test of which metric carries the information, and it favours the absolute value.