A Trial Finds Benefit, a Cohort Finds Harm: Antifibrotic Therapy in RA-ILD
A retired ironworker's new lung disease sits between a randomized trial that found benefit and a 2025 real-world study that found the opposite — the disagreement is about which kind of evidence to trust, not which drug exists.
J.K., a 66-year-old retired ironworker, has lived with seropositive rheumatoid arthritis for twelve years, his hands and knees carrying the marks of both the disease and four decades of the trade before it. His rheumatologist started a new departmental surveillance protocol screening every RA patient's lungs regardless of symptoms, and J.K.'s scan came back showing a UIP-pattern interstitial lung disease he had never noticed — no cough he'd mention, just the breathlessness he'd quietly attributed to being sixty-six. Methotrexate, which had controlled his joints reasonably well, was stopped over concern about its own pulmonary toxicity and replaced with a biologic; his joint disease has stayed quiet since. His lung disease has not: FVC has fallen from 74% to 66% predicted over the eight months since diagnosis, tracked by his new pulmonologist rather than found retrospectively.
The INBUILD trial's own diagnosis-subgroup analysis found nintedanib slowed FVC decline consistently whether the underlying disease was IPF or an autoimmune ILD like his — real, randomized evidence, though the trial wasn't powered to make a diagnosis-specific claim on its own. Set against that is a single-center prospective cohort of 199 CTD-ILD patients from West China Hospital, reported as an ATS 2025 conference abstract rather than a peer-reviewed paper, which found the opposite direction entirely: those started on an antifibrotic had significantly worse FVC at both twenty-four months (p=0.003) and thirty-six months (p=0.02) than those who weren't. Read side by side, one of these findings has to be explained rather than simply weighed against the other — and the same 2025 report notes, almost in passing, that its antifibrotic-treated group already had worse lung function and more UIP-pattern disease at baseline than the comparison group. The uncomfortable part is that the imbalance describes him. A UIP-pattern scan and eight months of documented decline are precisely what mark a patient out to be started on an antifibrotic in the first place, so J.K. would have sorted into that cohort's treated arm on exactly the features that made it look worse at baseline. Read one way that means the cohort's result is about patients like him and should worry the room; read the other it means the cohort measured his kind of patient's prognosis rather than his kind of patient's response to a drug. The second reading is the one INBUILD's randomization was built to protect.
Two papers, two directions
Start nintedanib. INBUILD randomized 663 patients with progressive fibrosing ILD, roughly a quarter of them with an autoimmune diagnosis, and its diagnosis-subgroup analysis found consistent FVC-decline benefit across that split. He already has eight months of documented decline of his own — that's the population this trial was built to inform, and randomization is what lets us trust the finding at all.
You're right that INBUILD's randomization makes it the more rigorous evidence in principle. But Wang, Xu, and Luo's 199-patient CTD-ILD cohort found the antifibrotic-treated group had significantly worse FVC at both 24 and 36 months — numbers describing patients managed the way J.K. is actually being managed, not a trial's enrollment window. I'll concede it's a conference abstract, not a full paper.
“More rigorous in principle” doesn't mean more relevant in practice when the rigorous trial's own follow-up ends well before the divergence the real-world data describes even becomes visible.
The 2025 study's own baseline table reports its antifibrotic-treated group already had worse lung function and more UIP-pattern disease before treatment even started — that's confounding by indication, sicker patients being treated preferentially, not proof the drug caused the difference. INBUILD's randomization protects against exactly that problem, which is why I'd still start nintedanib — but J.K.'s own trajectory from here should be tracked as real evidence, not assumed to match either paper.
Agreed: start nintedanib, with FVC and DLCO rechecked at three and six months rather than the usual annual interval, given the real disagreement about what his long-term trajectory might do.
Not agreed: how much weight the 2025 cohort study should carry in future CTD-ILD patients generally — the rheumatologist wants it treated as a genuine caution requiring closer surveillance sitewide, and expects it to read differently if it reaches full publication with the baseline imbalance adjusted for; the pulmonologist and clinical pharmacologist see it as adequately explained by that imbalance and not grounds to change practice on its own.