Effect of dose tapering policy on the prescription of advanced therapy for rheumatoid arthritis: a retrospective analysis of clinical and cost outcomes in Taiwan

In total, 9,099 adult patients with RA who had received aRA treatment for >3 months between January 2011 and October 2017 were included in the study (Supplementary Material 1). The mean age of patients at the start of follow-up was 57.2 years (SD 13.3), and 78.7% were female. The total number of follow-up periods was 239,789, equivalent to 59,947 patient-years (Table 1). Prescriptions for RA treatment and medical costs were aggregated every 3 months, along with the follow-up. On average, patients received a median of two csDMARDs, with a median dose of 80 DDDs of csDMARDs, 30 DDDs of NSAID and 180 mg of steroids every 3 months. The median cost for all-cause medication was 2983 USD; for RA-related medicine, the cost was 2636 USD; medication accounted for most of the medical costs (Table 2).

Table 3 shows the adjusted effect of the tapering policy and aRA treatment duration. We discovered that aRA dose tapering significantly increased when the treatment duration lasted for >24 months in the prepolicy period (aOR = 3.27, p < 0.001) (Hypothesis 0). After implementing the policy, this increase was further escalated to an aOR of 4.00 (p < 0.001). Comparing the aOR of postpolicy to that of prepolicy, it increased by 1.2-times, which was statistically significant (p = 0.012) (Hypothesis 1-1). Conversely, in treatment duration <24 months, the association between the implementation of the policy and the tapering of aRA dose was statistically insignificant (p = 0.16, aOR = 1.08) (Hypothesis 2-1).

Figure 2 shows the result of the aRA dose tapering policy on other nonadvanced RA treatments. The dose-tapering policy, which regulates the use of aRA, was also associated with the use of other RA treatments, including the dose of steroids, the number of csDMARDs, DDDs of csDMARD per day and DDDs of NSAID per day. Notably, all four factors were significantly reduced after the policy was implemented in 2014, with relative reductions ranging 2–4%. These were unintended effects of implementing aRA dose-tapering policy (Hypothesis 2-2). These reductions were statistically significant in both subgroups defined by aRA duration <24 months and ≥24 months, except for the use of steroids in the treatment ≥24 months sub-group, which increased by 1.04 (p < 0.01) and the test for policy by treatment duration interaction reached statistical significance (p < 0.01).

We also examined the effect of tapering aRA dose on non-aRA treatment (Hypothesis 1–2), which was considered the intended policy effect mediated indirectly through aRA treatment on those undergoing non-aRA treatments. A consistent reduction in non-aRA treatments was associated with tapering aRA dose; point estimates of aORs of all sub-groups ranged from 0.87 to 0.97, and no statistically significant tapering by policy interaction was observed (p = 0.08–0.86).

Figure 3 shows the result of the aRA dose tapering policy on costs. Implementing the aRA treatment dose-tapering policy was significantly associated with reduced medication costs without mediation through aRA treatment dose tapering (all: p < 0.01). This was considered an unintended effect of the policy (Hypotheses 2–3). These reductions were consistent for all causes and RA-related medication costs, and the relative reductions were at a consistent level of 7% in treatment reductions longer and shorter than 24 months. Tests for interaction were insignificant.

For hypotheses 1–3, the intended (mediated) effect of aRA tapering prescription on cost significantly reduced all-cause and RA-related medication costs. The relative reductions were 24–31% (all: p < 0.01). Despite the significant reduction in costs following policy implementation, the level of relative reduction decreased from approximately 30% (prepolicy subgroup) to 24% (postpolicy subgroup), with a test for interaction showing p < 0.01 for both all-cause and RA-related medication costs. The association between policy implementation and nonmedication costs was insignificant, and no significant interaction was observed across all subgroups.

The EULAR and ACR guidelines strongly recommend a treat-to-target treatment approach, which requires frequent monitoring of disease activity and treatment adjustment until remission or LDA is achieved [1,4]. The need to maintain long-term full-dose aRA till a predefined treatment target is reached has been a topic of debate. Once the inflammatory activity has been resolved, maintaining a minimum dose can effectively control the disease [1,22]. The concept of tapering aRA in RA treatment has been introduced in several guidelines and studies [1,4,23], and it may be necessary in clinical practice to enhance cost-effectiveness and minimize the risk of drug exposure [24,25], We aimed to provide data to evaluate the impact of dose tapering on clinical practice in a real-world setting.

To our knowledge, this study is the first nationwide investigation of the policy implications of aRA in patients with RA. In Taiwan, RA management is significantly influenced by the NHI reimbursement system [5]. In 2014, a dose tapering policy for aRA was introduced for RA treatment by the NHI, providing a unique opportunity for a natural experiment at a nationwide level with comprehensive coverage. Under the new reimbursement policy, prescription adjustments are enforced when the treatment duration reaches 24 months and patients are in a stable disease condition, subject to the clinical judgment of the prescribing physician; thus, not all patients will have their prescriptions reduced. In contrast, in the present study, we found that, even before the implementation of the reimbursement policy, dose tapering was commonly observed in patients who had received aRA for >24 months. The implementation of the tapering policy resulted in an additional but moderate increase in dose tapering, with the level of association increasing three to four-times, indicating a statistically significant difference. However, the policy implementation did not demonstrate a significant association with dose tapering prescriptions in patients who had received advanced therapy for <24 months. The observed prescription pattern in the real world aligned with the intention of the reimbursement policy.

The effect of the policy on clinical practice and cost outcomes is primarily an indirect consequence of changes in aRA prescriptions, which represents the intended effect of the policy. In our study, dose tapering, defined as a >50% reduction compared with the maximal dose for the same patient, was associated with reduced use of steroids, csDMARDs, and NSAIDs and lower total medical and medication costs. The change in aRA prescriptions did not significantly affect nonmedication costs in any scenario. These outcomes align with the policy's intent.

After implementing the tapering policy, we observed reduced medical costs associated with prescription changes. Before policy implementation, tapering aRA dose was associated with a 27% reduction in total medical costs; this reduction attenuated to 22% after the policy was implemented. The difference was statistically significant, suggesting a potential partial offset of the cost reduction resulting from tapering advanced therapy prescriptions in the context of the implemented policy.

This study referred to the effect of tapering policy on outcomes that bypass prescription changes as the unintended policy effect. We found that implementing the dose tapering policy resulted in a 7% reduction in steroid dose in patients with a treatment duration of <24 months. This reduction in steroid use was not within the intended scope of the policy. A possible explanation for this phenomenon is that the policy standardized the timing of aRA dose tapering. During the initial 24 months of treatment, physicians may have been more flexible in adjusting steroid use due to reduced concerns about the high cost of aRA and the need to maintain a suitable dose.

In contrast, implementing the policy was associated with a 4% increase in steroid dose among patients who received treatment for >24 months. This finding may reflect additional steroid use following aRA dose tapering. aRA dose tapering in accordance with policy requirements may have led to supplementary steroid use to maintain disease control. This pattern may reflect either a response to potential loss of disease control or a proactive adjustment in prescribing behavior to align with policy requirements. However, this interpretation remains limited given the absence of direct clinical outcome measures.

A similar trend of potential partial offset of the unintended policy effect was also observed, with a lesser reduction in the use of csDMARDs, NSAIDs and medication-related costs. The unintended policy effect on medication-related costs aligns with decreased cost reduction after policy implementation.

Our findings are consistent with those of several previous studies. For example, in the Dose Reduction Strategy of Subcutaneous (DRESS) tumor necrotic factor inhibitors (TNFi) trial [26], the dose reduction strategy was associated with cost reduction; however, a higher percentage of patients in the dose reduction group initiated csDMARDs or dose escalation and received intramuscular and intra-articular glucocorticoid injections than those in the usual care group. Similarly, in the Spacing of TNF-blocker injections in a Rheumatoid Arthritis Study trial, patients with RA in remission status who were administered TNFi at a stable dose for at least 6 months were randomly assigned to either continued maintenance or the TNFi injection spacing strategy [27]. The spacing strategy was associated with reduced costs but also a significantly high risk of disease relapse, which was defined as DAS28 >2.6 with an increase of >0.6 [28,29]. Another study comparing the clinical and economic impact of maintaining TNFi treatment versus dose tapering in patients with RA who achieved remission or LDA in real-world practice reported that dose tapering resulted in reduced healthcare-related costs but a shorter duration of sustained disease control [30]. These studies highlight the complex trade-off between cost reduction and maintaining disease control when implementing dose-tapering strategies in RA treatment.

Our study showed that implementing the NHI dose tapering policy in patients with LDA reduced healthcare costs and decreased the use of RA-related medications for most patients. However, a mandatory dose reduction can cause RA disease flare-ups in certain patients, which may be reflected by the increased steroid dose during the postpolicy phase. This increased steroid use may have counterbalanced the reduction of medication-related costs after policy implementation, possibly explaining why this reduction was less pronounced. These findings suggest that additional policies or strategies may be necessary to address the needs of patients with RA who may experience disease flare-ups postdose reduction.

The limitations of this study include the following. This was an observational study, and although we adjusted for some demographic and clinical factors to control potential confounding factors, other unmeasured confounding factors may still influence the associations related to tapering policies. We did not identify any strong confounding factors; however, using randomization to eliminate the influence of all potential confounding factors would be extremely challenging when evaluating the policy effects at a national scale. In addition, we did not perform an interrupted time-series analysis to explicitly model pre-policy trends and temporal changes; therefore, the observed findings may have been influenced by underlying secular trends and should be interpreted as associations rather than causal effects. The definition of dose tapering based on PDC may not fully distinguish intentional dose reduction from temporary treatment interruption or delayed prescription refills. In addition, this approach may not have fully captured clinical tapering strategies such as dose reduction or interval extension, and therefore may have missed more gradual tapering patterns. Changes in available aRA therapies over time may also have contributed to the observed changes in medication use and costs, despite adjustment for index therapy in the model. The application of multiple exclusion criteria may have affected the representativeness of the study population and limit the generalizability of the findings. Furthermore, the lack of direct measurement of RA disease severity, as indicated by the ACR criteria. Therefore, we relied on the use of RA medications other than aRA as an alternative measure. These measurements may not directly capture disease progression; however, there is a reasonable relationship between the two aspects, and medication utilization and costs were also relevant in understanding the impact of the policy. However, as these measures rely on proxy indicators rather than direct clinical outcomes, they may affect the estimation of the true clinical impact of the policy. The linkage of large-scale electronic medical records (EMRs) from multiple healthcare organizations, including NHIRD, is a developing trend, which may offer more direct information on clinical outcomes and will be necessary for further study.

In summary, before the advanced therapy tapering policy was implemented, there was a threefold increase in dose tapering prescriptions when the treatment duration reached 24 months. However, after the policy implementation, the dose tapering increased fourfold. Medication costs and other treatments for RA were associated with aRA dose tapering. However, our findings indicate a moderate partial offset of the cost reduction, reflecting both the intended and unintended effects of the policy. The results herein provide a comprehensive review of the impact of mandatory policies and outcomes of patients with RA in a real-world setting in Taiwan. At the time of this report, the tapering policy continues to be effective, and the findings of both the intended and unintended effects of the policy are expected to be upheld. Our findings could offer valuable insights for healthcare providers and policymakers of healthcare reimbursement when considering the use of aRA treatments.

K-J Li contributed to the study conception and design and provided clinical expertise in interpreting the findings. C-H Tang and C-L Chang were responsible for data acquisition, statistical analysis and interpretation of health policy implications. H-F Liao and P-J Yen contributed to data analysis, manuscript writing and project administration. W-Y Shau supervised the overall project, contributed to the study design and critically revised the manuscript. All authors reviewed and approved the final manuscript.

The authors thank Chia-Chien Chen, Chih-Yi Hsin, Chacun Lai and Lung-hung Kuo for assisting with data generation and valuable reviews.

This study was sponsored by Pfizer, Taiwan.

This study was sponsored by Pfizer. H-F Liao and P-J Yen are employees of Pfizer. The authors have no other competing interests or relevant affiliations with any organization or entity with the subject matter or materials discussed in the manuscript apart from those disclosed.

Editorial support and manuscript formatting support were provided by Anisha Kunder at Editage and were funded by Pfizer.

The study protocol was reviewed and exempted from approval by the Ethical Review Board of the National Taiwan University Hospital.

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

The authors certify that this manuscript reports the original results of a real-world evidence study. A prespecified study protocol was developed and is not publicly available. No preregistration was reported for this study. The underlying data and analytic code used in the analysis are not publicly available. Reporting checklists were not provided for this manuscript.

This work is licensed under the Attribution-NonCommercial-NoDerivatives 4.0 Unported License. To view a copy of this license, visit https://creativecommons.org/licenses/by-nc-nd/4.0/