All RA-ILD deserves a trial of immunomodulation Save

The optimum treatment approach in RA-ILD remains an area mired in controversy. This is evident from the contrasting views taken by the 2 recent expert guidelines from ACR and EULAR, which advocate for very different treatment options. 

The approach to RA-ILD with a UIP phenotype is even more controversial. This is driven by the radiologic similarities to idiopathic pulmonary fibrosis (IPF). Immunomodulation was previously the treatment of choice in IPF, based on expert opinion and non-RCT data, but all of that abruptly stopped when the PANTHER-IPF trial showed increased mortality with immunomodulation in IPF. It is a reasonable argument however that UIP RA-ILD and IPF while sharing similar radiologic end-points, arrive at that destination in very different ways and that may have important treatment implications. Ultimately, a lack of data is what drives diverse opinions, and we can hope that further research will clarify these controversies.

It is within that background milieu that Erik Way and colleagues performed the study presented in Abstract 2688 presented at Tuesday’s poster session on imaging in rheumatic diseases. aiming at predicting the response to immunomodulators in RA-ILD. This study included 132 individuals with RA-ILD from 5 clinical sites in the US. 

Patients were divided into responders and non-responders. Responders were defined as having stable/improved FVC at 12 months and survival ≥12 months. Baseline characteristics (and predictors of RA-ILD) were similar between the 2 groups. 

The nonresponder cohort had a higher proportion of radiologist-identified UIP (56% vs 36%, p=0.047). However, when the authors constructed logistic regression models utilising clinical and imaging variable to predict response to immunomodulators, UIP alone performed poorly, with an area under the curve (AUC) of 0.477. This is worse than a coin toss. Adding in further quantitative imaging features improved this to AUC of 0.57. The highest-performing model incorporated both imaging and clinical variables (prior methotrexate use, presence of gastroesophageal reflux, biologic sex, UIP and other quantitative imaging features) demonstrated an AUC of 0.629. At the end of the day, all of these models are poor predictors of response to immunomodulation.

This tells me that we don’t know what we thought we knew about which RA-ILD patients will respond to immunomodulation. In my own practice, I tend to use immunomodulation in all patients with RA-ILD (in the absence of contra-indications) and this largely validates that approach. I think until we discover more about how to predict which patients will respond, we simply do not have the evidence to support denying immunomodulation to any specific patient group. 

Commencing immunomodulation and assessing treatment response remains the optimum approach at present. Much as we loathe a trial-and-error philosophy, it appears that we are stuck with it, at least for now.