In recent years, treatment of RA has transitioned from symptom management to a treat-to-target paradigm [14], which has demonstrated superior disease outcomes in clinical trials. However, many patients with RA are still unable to achieve DAS28(CRP) treatment targets of < 2.6 or ≤ 3.2. No single biomarker has been identified that will reliably inform a management decision [15], and no systematic method has been established for determining the optimal therapeutic strategy for a patient based on quantification of molecular assays of RA disease. This absence is especially marked given the wide range of biologic and small-molecule DMARDs available to treat RA and the recent efforts in precision medicine to personalize treatments using increasing knowledge of the heterogeneous functions and interactions among immune cells across different patients [16].

The current analysis is the first to examine the association of baseline NLR with improvement of RA disease in a large cohort of > 3000 patients across the MTX-naïve and MTX-experienced populations. It is also the first analysis to investigate the association of the NLR with RA disease improvement in patients who have received a JAK inhibitor.

In this post hoc analysis, we explored the relationship of the NLR with likelihood of response to filgotinib, a preferential JAK1 inhibitor, based on the previously published cut point of 2.7. Our analysis demonstrates that no consistent differences were observed in response rates between NLR-High and NLR-Low control-arm patients, all of whom received MTX or other csDMARDs. In contrast, sustained improvements, across a broad range of clinical and PRO metrics, were observed among NLR-High patients treated with FIL200 + MTX/csDMARD relative to NLR-Low patients receiving similar treatment. While these results are consistent with our hypothesis that high baseline NLR is associated with improved clinical response to filgotinib, the precise mechanism by which JAK inhibitor therapy may benefit NLR-High patients is not known. Future studies should further evaluate this mechanism as well as whether the baseline NLR is associated with clinical responsiveness to additional RA treatments.

The clinical and PRO improvements among NLR-High compared with NLR-Low patients treated with FIL200 + MTX/csDMARD were further reflected by differences in the NNTs. While NLR-Low patients exhibited NNTs of 6.2 (MTX-naïve) and 5.3 (MTX-experienced), among NLR-High patients, the NNT was 3.2 (both groups). Together with the observed clinical and PRO improvements, these observations support the utility of the > 2.7 NLR cut point for classifying a patient population who may demonstrate stronger improvements with JAK inhibition therapy. Furthermore, as a differential white cell count is included as part of a routine laboratory workup prior to initiation of a new treatment, the NLR is a readily available predictive biomarker with the potential to enrich the cohort of individuals likely to respond in daily clinical practice. It thus presents a possible example of a biomarker that could be used to personalize treatment while not increasing testing burdens on patients and healthcare providers. However, it is important to note that patients treated with filgotinib achieved greater DAS28(CRP) improvements from baseline than did controls across both NLR subgroups and in both populations studied.

While a significant difference in DAS28(CRP) rates < 2.6 was observed between NLR-High and NLR-Low in MTX-experienced patients who received FIL200 + csDMARD, only a directionally consistent shift was observed among MTX-naïve patients. However, given the differences in continuous clinical and PRO metrics at other time points, it seems likely those improvements are real in the MTX-naïve population. Nevertheless, additional studies in MTX-naïve populations are warranted as well as the utility of the 2.7 NLR cut point for other approved therapies in RA. Further research into the biologic mechanisms underpinning the NLR in RA might shed light on potential benefits of early treatment.

This study has a few limitations that should be noted. As with other post hoc analyses, the results should be interpreted with caution because of the potential for increased risk of Type 1 errors, particularly in the multiple analyses of subgroups. Additional prospective studies should be carried out to confirm the findings from this exploratory analysis. Additionally, although the regression models control for baseline disease activity (as represented by significant covariates in Table 1), we cannot discount the possibility that changes might be attributed to a greater dynamic range for changes in disease activity in NLR-High patients. However, such an explanation seems unlikely given that similar improvements were not observed in NLR-High patients who received placebo or ADA.