Rheumatoid arthritis is a major global health problem, associated with a high burden of disease from both individual and societal perspectives [2]. Early RA treatment provides the best opportunities to achieve disease remission and long-term damage prevention [6, 7, 25]. The existence of a window of opportunity for better treatment outcomes with an upper limit situated between 3 to 6 months after disease onset is now widely accepted [4, 18, 21]. Widespread efforts to reduce delays in RA diagnosis and treatment are needed, but the accomplishments of such initiatives in developing countries remain unclear due to scarcity of data.

We observed a significant decrease in diagnostic and treatment delays in the last decades in Brazil (Figs. 1, 2), such as reported in other nations [26,27,28]. Before 1990, the median delay to RA diagnosis in Brazil was 24 months and to receive the first DMARD was 60 months. These delays have decreased ever since and converged to a median of 8 months for diagnosis and 11 months for treatment in the period 2011–2015 (Table 3). Despite all the improvements, these numbers were still higher than those reported in developed countries, where treatment delay has mostly been situated around 4–8 months from disease onset [28,29,30]. This finding reinforces the need for regional data in order to understand the status of early RA management from a global perspective; generalizations are not warranted.

The proportion of patients receiving early RA diagnosis and treatment increased along the studied period considering all defined windows (Tables 1, 2). In 2011–2015, more than 70% of the patients with RA were diagnosed and treated within 12 months of symptoms onset. Nevertheless, in the same period, only 36.3% received treatment within 6 months and 17.2% within 3 months of symptoms onset (Table 2). This shorter (3-month) window has been associated with the best outcomes [4, 31, 32]. Therefore, despite all the improvements, even recently a very early beginning of RA treatment (where the most promising results are expected) was still difficult to achieve in Brazil.

We noted increasing effect sizes (V values) for the association between the year of symptoms onset (arranged in time intervals) and the proportions of patients receiving early RA management across the 3, 6, and 12-month thresholds. This indicated that the magnitude of changes in diagnostic and treatment delays was not the same in these 3 windows. Instead, changes were greater around the 12-month threshold. This phenomenon can also be perceived with the increasing odds ratios for early treatment across the 3, 6, and 12-month windows when comparing patients whose symptoms initiated in the 2011–2015 period and those with symptoms onset before 1990. In other words, large delays (of more than 12 months) were more easily reduced, whereas achieving delays of less than 3–6 months (which are the ideal goals of early management) proved to be more difficult.

The mean diagnostic and treatment delay decreased progressively across the studied periods. Patients whose symptoms initiated in 2011–2015 (reference group) had significantly shorter delays in both diagnosis and treatment when compared to all other groups; the further removed into the past the period of symptoms onset, the greater the differences in comparison to the reference group (Table 4). These findings suggest a sustained decrease in mean delays throughout the assessed period, rather than a transient phenomenon at some particular occasion.

However, as noticed before, the sustained decrease in mean diagnostic and treatment delays in the last decades did not imply a uniform change across the different window thresholds. Rather, the decreases were influenced more by the progressive elimination of very large delays (greater effect sizes around the 12-month windows) than by the increments in frequency of short delays (smaller effect sizes around the 3-month widows). This finding indicates that measuring the proportions of patients being treated within the targeted windows could be more useful for policy adjustments than simply measuring changes in mean delays.

The main limitation of this study was that it relied to a great extent on the information provided directly by participants regarding the delays between symptoms onset and the diagnosis and treatment of RA. When feasible, diagnostic and treatment delays were ascertained using medical records, but for most participants these data were not retrievable from previous records. Confidence intervals (CI) for the mean delays were computed on a year-to-year basis (Figs. 1, 2), revealing greater imprecision (larger CI), as expected, around the older point estimates in comparison to the more recent ones. Despite this limitation, our findings of decreasing delays in RA management are consistent with those reported in other countries [26,27,28].