The effective threshold dose of etanercept in patients with methotrexate-resistant rheumatoid arthritis

Eligible patients’ data were extracted when the following inclusion criteria were met: (1) all patients fulfilled the 2010 criteria for RA of the American College of Rheumatology or European League Against Rheumatism [21]; (2) they had an inadequate response to initial MTX monotherapy, defined as DAS28-ESR > 3.2 (28-joint disease activity score using erythrocyte sedimentation rate) after receiving MTX for at least 6 months with a stable route of administration and more than 10 mg weekly for at least 3 months prior to the baseline visit [22]; (3) and subsequently using etanercept 25 mg twice weekly (BIW) by subcutaneous injection; (4) had at least 24 months of follow-up after giving etanercept. Patients were excluded if they were illiterate, had severe mental disorders, had serious physical constraints, or were over the age of 85. The tapering and increase of etanercept or switch to another biologic agent were allowed according to the T2T recommendations. Conventional synthetic disease-modifying antirheumatic drugs (csDMARDs), non-steroidal anti-inflammatory drugs (NSAIDs), and glucocorticoids (GCs) were allowed to administrate at stable doses if needed. Tapering of GCs or csDMARDs was allowed according to routine practice, but not to be changed within 2 weeks before assessment.

The cumulative dose for 24 months of each individual was recorded, and the effective cut-off of dose was calculated to identify the remission response (DAS28-ESR < 2.6) at 12 months [23]. The appropriate cut-off value of 12-month cumulative dose was estimated by constructing a receiver operating characteristics (ROC) curve. The Youden index is a measure of the accuracy of the model, which is used to determine the optimal cut-off value. Correlations were calculated by using ROC analyses and the point estimates of the area under the curve (AUC) and surrounding confidence interval (CI) to verify that the lower limit of CI was above 0.5. The sensitivity (true positive rate) represents the proportion of observations that were predicted to be positive when they were positive. In contrast, the 1-specificity (false positive rate) represents the proportion of observations that were predicted to be positive while they were negative.

The outcome assessment included the proportion of patients having reached at least one remission/LDA over 2 years, average daily exposure to glucocorticoid, the proportion of patients’ withdrawal from GCs, and cost-effectiveness by the available extended 2-year data in patients. DAS28 scores, laboratory values, quality of life assessments by EuroQol five-dimensions questionnaire (EQ5D), and medications during the follow-up were recorded [24]. The changes of DAS28 over 24-month follow-up were also analyzed. Drug safety was assessed by self-reported records among the overall population at each follow-up record. The occurrences of disease flare were defined as follows: (1) DAS28 > 3.2; (2) increase dose in DMARDs and/or GCs; (3) switch to another biologic agent.

We developed a decision tree model to assess the mean between-group difference in costs and quality-adjusted life years (QALYs) gained over 24 months among the patients who completed a 2-year follow-up without switching therapy regimens. Based on the Chinese time-trade-off method, the EQ-5D questionnaire responses yielded health utility values, from which we calculated cumulative QALYs [24]. An incremental cost-effectiveness ratio (ICER) was calculated by dividing the cost difference by the QALY difference in per pair of treatment schemes [26]. A 10,000 iteration Monte Carlo simulation was implemented to get the cost-effectiveness acceptability curve (CEAC) to analyze the preferred treatment group in a range of willingness-to-pay (WTP) threshold intervals. And the cost-effectiveness plane was also developed to view the distribution between each iteration and WTP threshold. The WTP thresholds under the per capita gross domestic product (GDP, around $12,552.08 in 2021) of China was considered cost-effective. All costs sourced from China in this study were converted into US dollars ($1 = RMB 6.4512, average exchange rate for 2021).

This observational cohort was established following the OPCSP trial (ClinicalTrials.gov ID: NCT03024307), a randomized trial on a compliance improvement program among rheumatic patients by a multidisciplinary team. The research protocol was approved by Shanghai Jiaotong University School of Medicine, Renji Hospital Ethics Committee (approval No. [2016] 216 K). All participating patients provided written informed consent.

Categorical variables were described by frequency and percentage, while continuous variables were described by mean ± standard deviation or median and quartile. The comparisons between the study groups were performed by Student’s t-test for continuous variables and the chi-square test (or Fisher’s exact test) for categorical variables. We also used restricted cubic splines (RCS) with four knots to flexibly model the association of predicted cumulative dose with sustained remission response [27]. RCS can identify non-linear relationships, provide more precise effect size estimates, facilitate thresholds identification, and improve predictive accuracy. The change of DAS28 over time was compared by specifying a linear mixed model with treatment and time as fixed factors and was performed with baseline DAS28 score as a covariate. As the visit schedules were slightly different between both groups, only baseline, 6-month, 12-month, and 24-month data were included for the analysis of the secondary outcomes. All statistical calculations were performed using the statistical software package IBM SPSS version 26.0 (IBM, Armonk, NY, USA). For all tests, p < 0.05 was considered to be significant.

Seventy-eight out of 549 patients of the cohort were eligible participants, and 72 patients completed 2-year follow-up. Baseline demographics and clinical characteristics of patients were summarized in Table 1. The average disease duration was 8.6 years and all of them had a diagnosis of RA longer than 1 year. The proportion of patients with positive ACPA and/or RF was over 80%. The average disease activity was 5.0 according to DAS28-ESR and 20 patients (27.7%) were in high disease activity. In terms of concomitant medication, the average dose of MTX was 10 mg weekly. Twenty-eight patients (38.8%) received GCs in this study at an average daily dose of 8.5 mg.

We used restricted cubic splines to flexibly model and visualize the dose–response relation of the predicted cumulative dose with sustained remission response without flare (as defined in the “Outcome assessment”) at 24 months in RA patients (Fig. 1A). The probability of the sustained remission response was relatively flat until around 1975 mg of the predicted cumulative dose and then started to increase rapidly afterward (p for non-linearity = 0.009). Above 1975 mg, the hazard ratio per standard deviation higher predicted cumulative dose was 1.00 (0.99 to 1.01). Seventy-two participants who completed 2-year follow-up were divided into saturated dose group (n = 36) and non-saturated dose group (n = 36) by the cut-off dose (Supplementary Figure S1).

The 6-month and 1-year cumulative cut-off doses of 1075 mg and 1625 mg for remission response were identified from the ROC curve; area under ROC was 0.613 (95%CI 0.472–0.754, p = 0.072) and 0.730 (95%CI 0.069–0.851, p = 0.001) (Supplementary Figure S2). Youden’s index was calculated as J = 0.195 and J = 0.369 with the sensitivity of 45.5% and 56.0%, and specificity of 74.0% and 80.9% respectively (Supplementary Table S1).

According to the cumulative cut-off dose, the recommended threshold dosing frequency of etanercept for the first 6 months, the next 6 months, and the second year were twice weekly (BIW), every week (QW), every 2 weeks (Q2W), and every month (QM) respectively (Fig. 1B).

A significant difference of the net changes of DAS28-ESR score over 2 years was observed, a greater average change in DAS28-ESR reduction in the saturated dose group than in the non-saturated dose group (average change 0.569, 95%CI 0.236–0.901, p = 0.001) (Fig. 2A). Greater net changes of DAS28-CRP score were also observed in the saturated dose group than in the non-saturated dose group (average change 0.425, 95%CI 0.133–0.717, p = 0.005) (Fig. 2B).

The proportion of patients achieving remission (27.8% vs 72.2%, p < 0.001) and LDA (58.3% vs 83.3%, p = 0.020) in the non-saturated group was both significantly lower than that in the saturated group at 24 months. In comparison, 16.5% of the non-saturated group achieved sustained LDA and 72.2% of those assigned to the saturated group had achieved sustained LDA at 24 months (p < 0.001). And the sustained remission rates in the non-saturated group were both significantly lower than those in the saturated group at 24 months (8.3% vs 52.8%, p < 0.001).

The remission (16.7% vs 52.8%, p = 0.001) and LDA rate (25.0% vs 75.0%, p < 0.001) in the non-saturated group were both significantly lower those than in the saturated group at 12 months. However, there were no significant differences between the two groups in remission and LDA rate at 6 months. At the endpoint of 24 months, among the 72 patients available for analysis, daily dose of GCs (4.0 vs 5.8 mg, p = 0.316) was not significantly (Table 2).

There were no serious adverse effects observed during the follow-up, and no difference in the proportion of patients experienced comorbidities (50.0% vs 63.9%, p = 0.234) or adverse effects (19.4% vs 27.8%, p = 0.405) across treatment groups (Table 3). The most common comorbidity in our study was cardiovascular diseases (30.6%), followed by digestive system diseases (23.6%), hematological system diseases (18.1%), osteoporosis (16.7%), and renal disease (16.7%). The two most common recorded adverse effects were infections (8.3%) and hepatotoxicity (8.3%), followed by gastrointestinal response (5.6%) and leukopenia (1.4%).

Medication costs and QALYs for each treatment group are presented in Table 4. The medication costs for 2 years in the ENT non-saturated dose group was $3320.66, and $4450.51 in the saturated dose group. The patients gained an average QALYs of 1.391 in the non-saturated group and 1.587 in the saturated group. The ICER yielded by the saturated dosing therapy was $5791.2 per QALY compared with the non-saturated dosing therapy, within the WTP threshold set as the per capita GDP of China.

A 10,000 iteration Monte Carlo simulation was conducted to evaluate the robustness of cost-effectiveness analysis. The output was displayed in the form of cost-effectiveness acceptable curve (CEAC) and cost-effectiveness plane (CE plane). The CEAC (Fig. 3) showed the saturated group was 69.3% of being cost-effective at the $12,552.08/QALY threshold. And the non-saturated group was 30.7% at the same threshold. The CE plane also revealed the results (Supplementary Figure S3). Each dot represents one output. Around 6930 outputs were below the WTP threshold in the saturated dose group. In the setting of 1 × GDP, the results of the saturated group showed high probability of being cost-effective regimen than the non-saturated dose group.