This study provides real-world data of RA patients initiating treatment with the first b/tsDMARDs, confirming that there are gender differences in clinical characteristics and timing of prescription.

In general, males had more comorbidities (e.g., higher BMI, Charlson comorbidity index, smoking, diabetes, …), with the exception of osteoporosis and Sjogren’s syndrome, which were more frequent in females. These findings are consistent with those previously described in the literature [5]. Additionally, disease onset differed between sexes, occurring at a significantly younger age in females. A previous study designed to estimate the incidence of RA in Spain also found a significantly lower age of RA onset in women [22].

In our study, factors associated with an earlier start of b/tsDMARDs were tobacco use, obesity, and concomitant treatment with methotrexate or glucocorticoids, the latter consistent with previous findings [23, 24]. In addition, treatment was initiated earlier in the periods 2007—2016, and ≥ 2017, compared to before 2007. This could be attributed to greater confidence in the use of b/tsDMARDs by prescribers, as well as greater availability of drugs with different mechanisms of action, in the more recent cohorts. Further research should explore the reasons behind the earlier prescription in patients with obesity and smoking habit. Assuming sex is “assigned at birth”, this attribute precedes other studied characteristics (e.g., smoking, BMI, comorbidities) among which sex differences have been observed. While BMI and smoking could act as potential mediators, they were not effect modifiers of sex (data not showed): sex had an effect in timing of b/tsDMARD start regardless of BMI and smoking habit.

Conversely, the initiation of the first b/tsDMARD occurred with a longer time course of RA in females, and among older patients. To our knowledge, this is the first study to show a gender difference in the timing of prescribing the first biologic agent, among those initiated after 2017. There is only one study that prospectively analyzed factors associated with RA disease duration at the time of prescription, and no gender differences were found [23]. It should be noted that patients in said study were included from 2003 until 2015 with a relatively low sample size (n = 178, compared to 900 in the corresponding cohorts in our study), and we just found significant gender differences in the most recent cohort (≥ 2017). The delay in initiating biologic therapy in elderly patients with chronic arthritis has been previously described [23,24,25,26] and may be explained by prescribers’ reluctance to initiate b/tsDMARDs in elderly patients.

At the start of the first b/tsDMARD, females had a higher activity index measured by DAS28-ESR, but there was no difference with males when using DAS28-CRP. In addition, ESR was significantly higher in females, while PCR was higher in males. The equivalence between DAS28-ESR and DAS28-CRP, established by Wells et al. [27], deserves some reflections: firstly, this equivalence was not assessed independently in men and women, and, secondly, when there were discrepancies in remission classification defined by DAS28, most were due to patients being classified as in remission with DAS28-CRP but not with DAS28-ESR, and the authors supported the criterion validity of DAS28-CRP. Therefore, it could be interpreted that DAS28-CRP is considered more reliable as a measure of disease activity than DAS28-ESR. ESR can be influenced by several factors unrelated to inflammation, such as age [28], gender [29] or plasma proteins. We found that Sjögren’s syndrome was significantly more frequent in women and, when fitting the linear regression model for DAS28 and its different components, the variable ‘Sjögren’s syndrome’ was only significant in the ESR model.

In our study, analysis of the individual components of DAS28 only showed significantly higher values in women in the subjective components TJC and PGA-VAS, but not in the objective component SJC. It is well known that women have a higher sensitivity to pain than men due not only to biological factors (e.g., a higher pain threshold due to testosterone in men, or the higher number of pain receptors in women with a different expression of these receptors), but also to gender-specific factors, such as the environment and social interactions, which vary between the sexes [30, 31]. Validated composite disease activity measures have been fundamental in recent decades to guide assessment of disease status and treatment response, but their limitations are well-recognized [32, 33]. Several published trials and studies have shown sex differences especially in subjective parameters: in the BARFOT study, higher DAS28 values in women were mainly dependent on higher pain scores [34]; in the Orencia and Rheumatoid Arthritis Registry, while there was no difference in response to abatacept between men and women, DAS28, TJC and PGA-VAS were consistently lower in men during follow-up [35]. In the QUEST-RA study, in patients with no SJC (no or very little clinical disease activity), gender differences were clinically and statistically significant in all other measures of the American College of Rheumatology core data set and in fatigue [7].

A possible explanation for the delay in treatment initiation in women despite a higher activity index (DAS28-ESR) could be that clinicians are aware of the predominance of subjective components such as pain or general patient assessment over components more directly related to inflammation, such as CRP or joint swelling [36]. Buch et al. [37], in a recent review, classified patients with persistent symptoms in the absence of objective inflammation as refractory non-inflammatory RA (RINRA). Misdiagnosis of persistent symptomatology as joint inflammation may lead to unnecessary treatment with b/tsDMARDs, or cycling though several therapies, but this increased perception of pain and discomfort clearly indicates an impaired quality of life that needs to be addressed.

Among the strengths of this study are the use of real-world data from BIOBADASER, a well-known registry with monitored patient information, including participating centers throughout the country. The nationwide coverage and the long-term study period encourage the generalizability of results among the Spanish population. Covariates included in the regression models were restricted to those available in the registry, therefore, the main limitation was the lack of data regarding other potential confounders that are not collected (e.g., hypergammaglobulinaemia, educational level or other social educational status, date of first symptoms) or difficult to measure (e.g., prescription bias) and that could have an impact in the disease activity measures or the timing of prescription. On the other hand, another important limitation could be that sex is recorded in BIOBADASER as a dichotomous biological variable (male, female) and not as self-reported gender (understood as the socially constructed norm: male, female, trans man, trans woman, non-binary, etc.). If the data had information on self-reported gender, a better approach to investigating the role of sex (biological, e.g. relevant to BMI) and gender (social, e.g. relevant to BMI and smoking) would be possible, which would serve to strengthen the study and advocate for better data collection in the sex and gender domains.