Sara Ganhão1, Raquel Lucas2, João Eurico Fonseca3, Maria José Santos4, Diana Rosa Gonçalves5, Nathalie Madeira6, Cândida Silva6, Eduardo Dourado7, Raquel Freitas8, Joana Rodrigues9, Soraia Azevedo9, Teresa Martins Rocha1, Raquel Miriam Ferreira1, Salomé Garcia1, Bruno Miguel Fernandes1, Ana Rita Prata10, Maura Couto11, Rita Pinheiro Torres12, Inês Cunha13, Lúcia Costa1, Miguel Bernardes14. 1. Centro Hospitalar e Universitário de São João. 2. Faculdade de Medicina da Universidade do Porto, Porto, Portugal ; Instituto de Saúde Pública, Universidade do Porto, Porto, Portugal. 3. Hospital de Santa Maria, Centro Hospitalar Lisboa Norte Centro Académico de Medicina de Lisboa, Lisboa, Portugal; Instituto de Medicina Molecular, Faculdade de Medicina, Universidade de Lisboa, Lisboa, Portugal. 4. Hospital Garcia de Orta, Almada, Portugal; Rheumatology Research Unit, Instituto de Medicina Molecular, Faculdade de Medicina, Universidade de Lisboa, Lisboa, Portugal. 5. Centro Hospitalar Entre o Douro e Vouga, Santa Maria da Feira, Portugal. 6. Instituto Português de Reumatologia, Lisboa, Portugal. 7. Hospital de Santa Maria, Centro Hospitalar Lisboa Norte, Centro Académico de Medicina de Lisboa, Lisboa, Portugal. 8. Hospital Garcia de Orta, Almada, Portugal. 9. Unidade Local de Saúde do Alto Minho, Ponte de Lima, Portugal. 10. Centro Hospitalar e Universitário de Coimbra, Coimbra, Portugal. 11. Centro Hospitalar Tondela-Viseu, Viseu, Portugal. 12. Centro Hospitalar Lisboa Ocidental | Hospital Egas Moniz, Lisboa, Portugal. 13. Centro Hospitalar do Baixo Vouga, Aveiro, Portugal. 14. Centro Hospitalar e Universitário de São João ; Faculdade de Medicina da Universidade do Porto, Porto, Portugal.
Abstract
BACKGROUND: Remission/ low disease activity (LDA) are the main treatment goals in rheumatoid arthritis (RA) patients. Two tools showing the ability to predict golimumab treatment outcomes in patients with RA were published. OBJECTIVES: To estimate the real-world accuracy of two quantitative tools created to predict RA remission and low disease activity. METHODS: Multicenter, observational study, using data from the Rheumatic Diseases Portuguese Register (Reuma.pt), including biologic naïve RA patients who started an anti-TNF as first-line biologic and with at least 6 months of follow-up. The accuracy of two matrices tools was assessed by likelihood-ratios (LR), sensitivity (SN), specificity (SP), positive predictive value (PPV), negative predictive value (NPV) and area under the ROC curve (AUC). RESULTS: 674 RA patients under first-line anti-TNF (266 etanercept, 186 infliximab, 131 adalimumab, 85 golimumab, 6 certolizumab pegol) were included. The median (IQR) age was 53.4 (44.7-61.1) years and the median disease duration was 7.7 (3.7-14.6) years. The majority were female (72%). Most patients were RF and/or ACPA positive (75.5%) and had erosive disease (54.9%); 58.6% had comorbidities. At 6-months, 157 (23.3%) patients achieved remission (DAS28 ESR < 2.6) and 269 (39.9%) LDA (DAS28 ESR ≤ 3.2). Area under the curve for remission in this real-world sample was 0.756 [IC 95% (0.713-0.799)] and for LDA was 0.724 [IC 95% (0.686 -0.763)]. The highest LR (8.23) for remission state was obtained at a cut-off ≥ 67%, with high specificity (SP) (99.6%) but low sensitivity (SN) (3.2%). A better balance of SN and SP (65.6% and 73.9%, respectively) was observed for a cut-off >30%, with a LR of 2.51, PPV of 43.3% and NPV of 87.6%. CONCLUSION: In this population, the accuracy of the prediction tool was good for remission and LDA. Our results corroborate the idea that these matrix tools could be helpful to select patients for anti-TNF therapy.
BACKGROUND: Remission/ low disease activity (LDA) are the main treatment goals in rheumatoid arthritis (RA) patients. Two tools showing the ability to predict golimumab treatment outcomes in patients with RA were published. OBJECTIVES: To estimate the real-world accuracy of two quantitative tools created to predict RA remission and low disease activity. METHODS: Multicenter, observational study, using data from the Rheumatic Diseases Portuguese Register (Reuma.pt), including biologic naïve RApatients who started an anti-TNF as first-line biologic and with at least 6 months of follow-up. The accuracy of two matrices tools was assessed by likelihood-ratios (LR), sensitivity (SN), specificity (SP), positive predictive value (PPV), negative predictive value (NPV) and area under the ROC curve (AUC). RESULTS: 674 RApatients under first-line anti-TNF (266 etanercept, 186 infliximab, 131 adalimumab, 85 golimumab, 6 certolizumabpegol) were included. The median (IQR) age was 53.4 (44.7-61.1) years and the median disease duration was 7.7 (3.7-14.6) years. The majority were female (72%). Most patients were RF and/or ACPA positive (75.5%) and had erosive disease (54.9%); 58.6% had comorbidities. At 6-months, 157 (23.3%) patients achieved remission (DAS28 ESR < 2.6) and 269 (39.9%) LDA (DAS28 ESR ≤ 3.2). Area under the curve for remission in this real-world sample was 0.756 [IC 95% (0.713-0.799)] and for LDA was 0.724 [IC 95% (0.686 -0.763)]. The highest LR (8.23) for remission state was obtained at a cut-off ≥ 67%, with high specificity (SP) (99.6%) but low sensitivity (SN) (3.2%). A better balance of SN and SP (65.6% and 73.9%, respectively) was observed for a cut-off >30%, with a LR of 2.51, PPV of 43.3% and NPV of 87.6%. CONCLUSION: In this population, the accuracy of the prediction tool was good for remission and LDA. Our results corroborate the idea that these matrix tools could be helpful to select patients for anti-TNF therapy.