RESEARCH ARTICLE The comparative responsiveness of Hospital Universitario Princesa Index and other composite indices for assessing rheumatoid arthritis activity Isidoro González-Álvaro ID1*, Isabel Castrejón2, Loreto Carmona3, on behalf of ACT-RAY, PROAR and EMECAR study groups¶ a1111111111 1 Rheumatology Service, Hospital Universitario La Princesa, IIS-IP, Madrid, Spain, 2 Division of a1111111111 Rheumatology, Rush University Medical Center, Chicago, Illinois, United States of America, 3 Instituto de a1111111111 Salud Musculoesquelética (InMusc), Madrid, Spain a1111111111 ¶ See Acknowledgments section for information about members, departments, institutions and countries a1111111111 * isidoro.ga@ser.es Abstract OPEN ACCESS Citation: González-Álvaro I, Castrejón I, Carmona L, on behalf of ACT-RAY, PROAR and EMECAR study Objective groups (2019) The comparative responsiveness of Hospital Universitario Princesa Index and other To evaluate the responsiveness in terms of correlation of the Hospital Universitario La Prin- composite indices for assessing rheumatoid cesa Index (HUPI) comparatively to the traditional composite indices used to assess dis- arthritis activity. PLoS ONE 14(4): e0214717. ease activity in rheumatoid arthritis (RA), and to compare the performance of HUPI-based https://doi.org/10.1371/journal.pone.0214717 response criteria with that of the EULAR response criteria. Editor: Feng Pan, University of Tasmania, AUSTRALIA Received: August 14, 2018 Methods Accepted: March 11, 2019 Secondary data analysis from the following studies: ACT-RAY (clinical trial), PROAR (early RA cohort) and EMECAR (pre-biologic era long term RA cohort). Responsiveness Published: April 10, 2019 was evaluated by: 1) comparing change from baseline (Δ) of HUPI with Δ in other scores Copyright: © 2019 González-Álvaro et al. This is an by calculating correlation coefficients; 2) calculating standardised effect sizes. The accu- open access article distributed under the terms of the Creative Commons Attribution License, which racy of response by HUPI and by EULAR criteria was analyzed using linear regressions permits unrestricted use, distribution, and in which the dependent variable was change in global assessment by physician (ΔGDA- reproduction in any medium, provided the original Phy). author and source are credited. Data Availability Statement: The authors would like to confirm that the data from the EMECAR and Results PROAR are the minimal data set supporting the ΔHUPI correlation with change in all other indices ranged from 0.387 to 0.791); HUPI’s stan- results of this study. These data are available as Supplementary material in 4 Excel files (EMECAR dardized effect size was larger than those from the other indices in each database used. In data wide, EMECAR data long, PROAR data wide ACT-RAY, depending on visit, between 65 and 80% of patients were equally classified by and PROAR data long). Further inquiries about this HUPI and EULAR response criteria. However, HUPI criteria were slightly more stringent, information may be directed to the Research Unit with higher percentage of patients classified as non-responder, especially at early visits. of SER at the following e-mail address: proyectos@ser.es. The data from ACT-RAY were HUPI response criteria showed a slightly higher accuracy than EULAR response criteria provided by Hoffmann-La Roche Ltd through a when using ΔGDA-Phy as gold standard. PLOS ONE | https://doi.org/10.1371/journal.pone.0214717 April 10, 2019 1 / 19 Responsiveness of HUPI data sharing agreement that does not allow for the Conclusion public sharing of these data. The authors did not enjoy any special access privileges in gaining HUPI shows good responsiveness in terms of correlation in each studied scenario (clinical access to these data. Regarding the possibility that trial, early RA cohort, and established RA cohort). Response criteria by HUPI seem more any other researcher would like to request data to stringent than EULAR’s. replicate the reported study findings, Hoffmann-La Roche Ltd has implemented a Data Sharing policy to align with the ICMJE recommendations: “Qualified researchers may request access to individual patient level data through the clinical study data request platform (www. clinicalstudydatarequest.com). Further details on Roche’s criteria for eligible studies are available Introduction here (https://clinicalstudydatarequest.com/Study- Sponsors/Study-Sponsors-Roche.aspx). For Objective evaluation of disease activity in rheumatoid arthritis (RA) has become a keystone of further details on Roche’s Global Policy on the disease management. Composite indices measuring disease activity have allowed implement- Sharing of Clinical Information and how to request ing treat-to-target and tight-control strategies, both contributing the most to the improvement access to related clinical study documents, see in RA outcome achieved in the last 15 years. The most frequently used indices to evaluate dis- here (https://www.roche.com/research_and_ development/who_we_are_how_we_work/clinical_ ease activity among rheumatologists are the Disease Activity Score (DAS28) [calculated with trials/our_commitment_to_data_sharing.html)”. C-reactive protein (CRP) or with sedimentation rate (ESR)] and the Simplified Disease Activ- ity Index (SDAI), since they have been widely validated, are endorsed by ACR and EULAR, Funding: Our study was supported by grants RD16/0012/0011 and PI14/00442 from the and are commonly used to assess therapeutic response in clinical trials [1–5]. In addition, the Ministerio de Economia y Competitividad (Instituto Clinical Disease Activity Index (CDAI) is being increasingly used, as it is easier to calculate de Salud Carlos III; Spain) and cofunded by the than the previous ones, despite limited validation. Fondo Europeo de Desarrollo Regional (FEDER). However, during the last 10 years, a fair amount of evidence suggested that both, DAS28 Data from ACT-RAY clinical trial were kindly and SDAI, present a gender bias, derived from differences between men and women in terms provided by Hoffmann-La Roche Ltd. No financial of pain perception and levels of erythrocyte sedimentation rate (ESR)[6–10]. Using these indi- support was received from Hoffmann-La Roche Ltd Data from EMECAR and PROAR cohorts were ces, the implementation of T2T strategy would be biased, leading to over-treatment in women, provided by the Spanish Society of Rheumatology. or under-treatment of men. This may lead to excess risk of adverse events in women or lower No financial support was received from the odds to achieve real disease control in men. In addition, assessment of response to treatment Spanish Society of Rheumatology. None of these in clinical trials might also be biased [11]. institutions played any role in the analysis or The Hospital Universitario La Princesa Index (HUPI) was developed to avoid a gender bias interpretation of data, nor were they involved in the writing of the manuscript. Roche and Sociedad in the assessment of RA disease activity by adjusting the contribution of tender joint counts Española de Reumatologı́a were involved in the and ESR by sex [12]. An additional advantage of HUPI is that it can be calculated with ESR, collection of data from ACT-RAY, and EMECAR and CRP, or both acute phase reactants (APR), producing an almost identical score and avoiding PROAR, respectively. However, these funders had missing data in longitudinal studies [12, 13]. no role in study design, analysis, decision to HUPI was initially developed and validated in PEARL, a longitudinal observational study publish, or preparation of the manuscript. nested in an early arthritis register, and is calculated as the sum of four variables (graded 0–3 Competing interests: Dr. Gonzalez-Alvaro reports according to their quartile distribution in the PEARL population [see S1 Table]): 28 tender grants from Instituto de Salud Carlos III, during the and swollen joint counts, global disease assessment by patient and APR [13]. When both ESR conduct of the study; consulting and speaking fees and CRP are used to calculate the index, the average of their scores in S1 Table is used to calcu- less than $10,000 from Lilly, non-financial support from UCB, consulting and speaking fees and non- late HUPI. Thus, the index ranges from 0 to 12, and its cut-offs for remission/low disease activ- financial support less than $10,000 from BMS, ity, low/moderate and moderate/high disease activity are 2, 5 and 9 respectively [13]. HUPI non-financial support less than $10,000 from may have a “ceiling effect”, especially in groups of patients with very high disease activity, such Pfizer, speaking fees less than $10,000 from as those included in clinical trials. Roche, speaking fees and non-financial support The objective of this study was to evaluate the responsiveness of the HUPI, in parallel to less than $10,000 from Abbvie, non-financial support less than $10,000 from MSD, outside the that of the classical indices—DAS28-ESR, DAS28-CRP, SDAI and CDAI—, in terms of corre- submitted work during the last 5 years; In addition, lation, using data from three cohorts. Furthermore, the performance of HUPI-based response Dr. Gonzalez-Alvaro has a patent PCT/ES2015/ criteria was compared with that of EULAR response criteria. 070182 issued. The remaining authors have no potential conflict of interest with regard to the PLOS ONE | https://doi.org/10.1371/journal.pone.0214717 April 10, 2019 2 / 19 Responsiveness of HUPI manuscript. This does not alter our adherence to Patients and methods PLOS ONE policies on sharing data and materials. Patients As mentioned, HUPI was developed in PEARL, an early arthritis register from Madrid (Spain) [12]. In the present study, we evaluate it in other RA populations, including RA from other countries. HUPI was evaluated in three different scenarios: a) an international clinical trial, the ACT-RAY—very high disease activity at baseline, homogeneous intervention, strict follow-up and patients enrolled in different countries; b) an early arthritis population (PROAR) in which sensitivity to change may be tested in a setting different from the early arthritis population used to develop and validate HUPI; and c) a long term prevalent RA population engaged in the pre-biologic era (EMECAR). The ACT-RAY clinical trial. ACT-RAY is a 2-year double-blind clinical trial (NCT00810199, EudraCT No 2008-001847-20) designed to evaluate the efficacy and safety of tocilizumab plus methotrexate or tocilizumab plus placebo in patients with persistent active disease despite methotrexate monotherapy. Inclusion criteria for ACT-RAY were RA classifi- cation according to 1987 ACR criteria [14], DAS28>4.4, and erosive disease, as described pre- viously [15]. Data collected included demographics, RA characteristics, as well as baseline and 4-weekly clinical and laboratory data necessary to calculate DAS28-ESR, DAS28-CRP, SDAI, CDAI and HUPI [15, 16]. Since no relevant differences in clinical response were reported between patients treated with tocilizumab in monotherapy or in combination with methotrexate [15], we included patients’ data irrespective of their allocation group. Considering that after the first year, patients in ACT-RAY were allocated into four different T2T strategies [16], for the present study the analysis was performed only with data from baseline and 12, 24 and 52 weeks visits. The PROAR cohort. PROAR was a longitudinal multicenter study including 5 consecu- tive patients from 34 Rheumatology Units in Spain. Patients were included if presented at least one swollen joint for less than a year, irrespective of fulfilling 1987 ACR criteria [14]. At base- line, patients should be naïve for disease modifying anti-rheumatoid drugs (DMARDs) or glu- cocorticosteroids. Evidence of infectious arthritis or crystal arthritis were considered exclusion criteria [17]. Follow-up was 5 years, from January 2001 to December 2006. Data collection included all variables needed to calculate DAS28-ESR, DAS28-CRP, SDAI, CDAI and HUPI at baseline and at each 6-monthly visits [17]. For the present study only patients fulfilling the 1987 ACR RA criteria along the follow-up were included. Most of these patients started treatment with DMARDs at the beginning of follow-up (S2 Table). Therefore, for the responsiveness analysis, baseline and 6 months visits were analyzed. The EMECAR cohort. EMECAR was a prospective longitudinal cohort of prevalent RA patients fulfilling 1987 ACR criteria [14] selected by random sampling in 34 Rheumatology Units from Spain. Follow-up took place from November 1999 to December 2004 with yearly visits. EMECAR database includes the required variables to calculate DAS28-ESR and HUPI, but not DAS28-CRP, SDAI or CDAI, since global disease assessment by physician (GDA-Phy) was not collected and C-reactive protein (CRP) values provided limited reliability. A detailed description of the EMECAR cohort has been published previously [18]. At baseline, no patient was under treatment with a TNF-antagonist or leflunomide. During 4 years of follow-up, 27% of patient started, at least, one of these treatments. As we have previ- ously described, improvement along the follow-up in this long term RA population was lim- ited [18]. However, since HUPI was developed in patients with early arthritis, we included information about EMECAR in order to compare the performance of HUPI compared to DAS28 in long standing disease. PLOS ONE | https://doi.org/10.1371/journal.pone.0214717 April 10, 2019 3 / 19 Responsiveness of HUPI Ethical statement This is a secondary analysis of anonymized data from patients included in ACT-RAY, EME- CAR and PROAR studies. ACT-RAY clinical trial was approved by the Research Ethical Com- mittee (REC) of all centers included in the study (see Acknoledgement section “Group ACT-RAY”). EMECAR study was approved by the REC of Hospital Universitario La Princesa and this approval was accepted by all centers included in the study (see Acknoledgement sec- tion “Group EMECAR”). PROAR study was approved by the REC of Hospital Universitario La Princesa and this approval was accepted by all centers included in the study (see Acknol- edgement section “Group PROAR”). ACT-RAY, PROAR and EMECAR studies were conducted according to the principles expressed in the Helsinki Declaration of 1983. All patients signed the respective written con- sent before study entry [15, 17, 18]. Statistical analysis We used STATA 12.0 for Windows (StataCorp LP, College Station, TX). To describe the three populations, means and standard deviation (SD), medians and interquartile range (IQR), as well as absolute and relative frequencies were used, depending on the distribution of variables. The external responsiveness of HUPI was evaluated as recommended by Husted et al in three different populations[19]. We used Pearson correlation coefficient to describe how changes from baseline (ΔHUPI) to different follow-up visits (ACT-RAY visits 12, 24 or 52 weeks; PROAR visit 6 months; EMECAR visit 4 years) correlated with corresponding changes in the values of global disease activity assessed by patient (ΔGDA-Pat), ΔGDA-Phy, ΔDA- S28-ESR, ΔDAS28-CRP. Spearman correlation was used with ΔSDAI and ΔCDAI, since the values of these indices do not follow a normal distribution. Internal responsiveness was also evaluated using standardized effect size (SES) calculated with MS Excell 2007 for Windows as the mean difference between baseline and each previously mentioned time points divided by the pooled standard deviation, as described by Hedges and Olkin [20, 21]. We evaluated how HUPI-based response criteria[13] behave in comparison to EULAR response criteria[22] using data from ACT-RAY. First, we tabulated the response with each set of response criteria and cross-tabulated them. To determine the accuracy of both response cri- teria, we used the percentage of correctly classified patients from the best fitted models with ΔGDA-Phy as external criterion. ΔGDA-Phy was used to avoid circularity, since neither HUPI nor DAS28 include this variable in their computation. Linear regression models using general- ized linear solutions (Stata’s glm command with the default option) were performed with ΔGDA-Phy (from baseline to different time points) as dependent variable and HUPI-based and EULAR response criteria as categorical variables. Beta coefficients with 95% confidence intervals (95%CI) for “Moderate” and “Good response” by either definition were reported, along with the Akaike information criteria (AIC) from each model (S6 and S7 Tables). The later allow us to identify the best model; given two different regression models fitted on the same data, the model with the smallest AIC value is considered the best [23]. Results Assessment of disease activity with different indices in three different populations Table 1 shows a description of the three study populations. In all three, about 75% of patients were women and mean age at baseline ranged from 53 to 61 years. As part of the inclusion cri- teria, patients from the early arthritis cohort had the disease for less than a year in contrast PLOS ONE | https://doi.org/10.1371/journal.pone.0214717 April 10, 2019 4 / 19 Responsiveness of HUPI Table 1. Baseline characteristics of the populations. ACT-RAY (n = 556) PROAR (n = 160) EMECAR (n = 789) Female gender, n (%) 446 (80.0) 113 (70.6) 568 (72.0) Age, median [IQR] 54 [46–62.5] 54 [44–68] 64 [54–71] Disease duration, median [IQR] (years) 8.3±8.2� 0.37 [0.21–0.63] 8.7 [4.1–12.6] Smoker, n (%) 200 (36.0) 60 (37.5) 274 (35.4) RF positive, n (%) N.A. 47 (34.6) 578 (74.6) ACPA positive, n (%) N.A. 76 (47.8) N.A. HAQ, median [IQR] 1.5 [1–1.875] 0.5 [0–1.125] 1.125 [0.5–1.875] HUPI, mean ± SD 10.9 ± 1.3 5.3 ± 3.1 6.5 ± 3.0 DAS28, mean ± SD 6.34 ± 0.99 3.84 ± 1.45 4.25 ± 1.40 DAS28-CRP, mean ± SD 6.44 ± 1.03 3.59 ± 1.33 N.A. SDAI, median [IQR] 44.3 [34.1–55.5] 13.9 [6.0–21.7] N.A. CDAI, median [IQR] 38.0 [28.8–47.2] 11.9 [4.6–20.4] N.A. Abbreviations: n, number; IQR, interquartile range; SD, standard deviation; DO, disease onset; N.A., not available; RF, rheumatoid factor; ACPA anti-citrullinated protein antibodies; HAQ, health assessment questionnaire; HUPI, Hospital Universitario La Princesa Index; DAS28, disease activity score calculated with erythrocyte sedimentation rate and 28 joint counts; DAS28-CRP, disease activity score calculated with C-reactive protein and 28 joint counts; SDAI, simplified disease activity index; CDAI, clinical disease activity index. � IQR of disease duration was not available for ACT-RAY patients. https://doi.org/10.1371/journal.pone.0214717.t001 with about 8 years in ACT-RAY and EMECAR. As expected, patients from the clinical trial showed the highest baseline disease activity and disability, EMECAR patients showed mid val- ues, and those from PROAR showed the lowest scores of disease activity and disability (Table 1). As a result of HUPI allowing calculation from CRP or ESR, whichever available at the study visit—a strategy to minimize missing data—, the HUPI was calculated in more visits than the other indices in the three populations: 99.8% of visits in ACT-RAY; 96.7% in PROAR; and 92.3% in EMECAR, with the only exception of CDAI in PROAR: 98.6% (S3 Table). In patients from ACT-RAY, baseline HUPI values show a “ceiling effect” with more than 40% of patients at the highest score of the index (12 units; upper left panel in Fig 1). The remaining indices did not show this effect, with less of 5% at the highest value of SDAI and no patient at the highest score of DAS28-ESR, DAS28-CRP and CDAI (remaining panels in Fig 1). All indices showed improvement of disease activity after starting treatment with tocili- zumab (Fig 1). Patients from PROAR and EMECAR showed lower disease activity at baseline, so no “ceil- ing effect” was observed in HUPI (Fig 1 left panels of mid and lower row). Disease activity improvement with all indices was observed in PROAR after starting DMARD treatment in this early arthritis population (Fig 1 mid row). Limited improvement was observed in EME- CAR (Fig 1 lower row). Responsiveness of HUPI versus traditional indices Despite its baseline “ceiling effect” in the ACT-RAY clinical trial, the change in HUPI score at week 12 had a good correlation with ΔGDA-Pat and slightly lower with ΔGDA-Phy. Conse- quently, the correlation was very good with ΔDAS28 either with ESR or CRP, and slightly lower with ΔSDAI or ΔCDAI (Table 2 and Fig 2). In ACT-RAY correlations tended to improve when comparing ΔHUPI from baseline and weeks 24 and 52 and the corresponding changes of the other variables (Table 2). Interestingly, PLOS ONE | https://doi.org/10.1371/journal.pone.0214717 April 10, 2019 5 / 19 Responsiveness of HUPI Fig 1. Evolution of disease activity during follow-up of the ACT-RAY, PROAR and EMECAR patients assessed with different indices. Data are presented as interquartile range (p75 upper edge, p25 lower edge, p50 midline in the box), p95 (line above the box) and p5 (line below the box). Dots represent outliers. https://doi.org/10.1371/journal.pone.0214717.g001 Fig 2 shows that there were individual important disparities (red dots) between ΔHUPI and ΔSDAI or ΔCDAI. We hypothesized that this correlation would not be perfect since HUPI was specifically developed to avoid gender bias of DAS28 and SDAI. In addition, HUPI does not include Table 2. External responsiveness of HUPI compared with improvement measured with other variables commonly used to assess response in rheumatoid arthritis. ACT-RAY PROAR EMECAR w12 w24 w52 w24 w52 Y2 Y4 (n = 491) (n = 475) (n = 340) (n = 136) (n = 129) (n = 560) (n = 440) ΔGDA-Pat 0.50 0.58 0.62 0.60 0.66 0.51 0.51 (0.43–0.56) (0.52–0.63) (0.55–0.68) (0.48–0.70) (0.55–0.75) (0.45–0.57) (0.44–0.78) ΔGDA-Phy 0.39 0.52 0.53 0.65 0.63 (0.31–0.46) (0.44–0.57) (0.45–0.60) (0.54–0.74) (0.51–0.72) ΔDAS28_ESR 0.77 0.78 0.78 0.91 0.89 0.87 0.87 (0.73–0.80) (0.74–0.81) (0.73–0.82) (0.88–0.94) (0.85–0.92) (0.85–0.89) (0.84–0.89) ΔDAS28-CRP 0.75 0.79 0.79 0.89 0.90 (0.71–0.79) (0.75–0.81) (0.75–0.83) (0.85–0.92) (0.86–0.93) ΔSDAI 0.52 0.56 0.51 0.75 0.78 (0.45–0.58) (0.49–0.62) (0.43–0.58) (0.67–0.82) (0.70–0.84) ΔCDAI 0.48 0.55 0.48 0.78 0.79 (0.41–0.55) (0.48–0.61) (0.39–0.56) (0.70–0.84) (0.72–0.85) GDA: global disease assessment; Pat: patient; Phy: physician; HUPI: Hospital Universitario La Princesa Index; DAS28_ESR or _CRP: disease activity score calculated with 28 joint counts and erythrosedimentation rate or C reactive protein; SDAI: simplified disease activity index; CDAI: clinical disease activity index; Data are shown as the correlation coefficient (95% confidence interval). https://doi.org/10.1371/journal.pone.0214717.t002 PLOS ONE | https://doi.org/10.1371/journal.pone.0214717 April 10, 2019 6 / 19 Responsiveness of HUPI Fig 2. External responsiveness of HUPI (ΔHUPI) compared with improvement measured with other variables commonly used to assess clinical improvement in rheumatoid arthritis. Red dots represent values in which there was clear discordance between ΔHUPI and other measuments of improvement. The correlation coefficients are shown at Table 2. https://doi.org/10.1371/journal.pone.0214717.g002 GDA-Phy in its calculations. So, in order to be able to compare the respective sensitivity to change (Internal responsiveness), we calculated the SES for each variable at the three time points studied. The SES for HUPI was always the highest in the three populations at all times studied (Fig 3 and S4 Table). In addition, the 95% CI of HUPI’s SES did not overlap with those from GDA-Pat, GDA-Phy, SDAI and CDAI at any time in ACT-RAY (Fig 3A and S4 Table). Similar findings were observed in patients from PROAR, although lower SES were observed since the baseline disease activity was lower than that of patients in ACT-RAY (Fig 1), with no significant differences across indices (Fig 3B and S4 Table). On the other hand, limited disease activity improvement had been described with DAS28 in EMECAR[18] and the data with HUPI were consistent with these previous findings (Fig 3C and S4 Table). Comparison of EULAR response criteria and HUPI-based response criteria In ACT-RAY, it was possible to determine HUPI response in more patients than with EULAR response criteria either at week 12 (528 vs. 518), week 24 (509 vs. 503) or week 52 (418 vs. 412). The lower number of assessments with EULAR response was due to missing ESR, required to calculate DAS28. In addition, the proportion of patients with no response was higher with HUPI than with EULAR response criteria, although gradually being the proportions closer PLOS ONE | https://doi.org/10.1371/journal.pone.0214717 April 10, 2019 7 / 19 Responsiveness of HUPI Fig 3. Internal responsiveness of different indices and variables used to assess disease activity in different cohorts. Data are shown as the standardized effect size (black circle) and its 95% confidence interval (black bar) at A) week 52 for ACT-RAY patients; B) week 24 for PROAR patients; and C) year 4 for EMECAR patients. https://doi.org/10.1371/journal.pone.0214717.g003 along the follow-up (Fig 4A and 4B and S1 Fig panel A). Similar findings were observed in PROAR and EMECAR (S2 Fig). Table 3 shows that response to treatment was equally classified by both criteria in > 65% of patients at week 12 and it gradually increased to >80% of patients at week 52. Three patients were classified as good responders with EULAR response criteria and non-responders with HUPI at all three time-points, whereas 1 patient was classified as good responder with HUPI but non-responder with EULAR only at week 12. The characteristics of these patients are shown in S5 Table. In summary, 3 patients had high number of tender joints, but low number PLOS ONE | https://doi.org/10.1371/journal.pone.0214717 April 10, 2019 8 / 19 Responsiveness of HUPI Fig 4. Comparison of EULAR and HUPI response criteria from baseline at different weeks of patients from ACT-RAY. A) Percentage of patients getting none, moderate or good response at week 12. B) Percentage of patients getting none, moderate or good response at week 52 (See S1 Fig panel A for information at week 24). C) Association of change in global disease assessment by physician (GDA-Phy) with the different categories of EULAR response (See S1 Fig panel B for information at weeks 12 and 24). D) Association of change in GDA-Phy with the different categories of HUPI response (See S1 Fig panel B for information at weeks 12 and 24). Data in panels C and D are shown as the predicted mean change in GDA-Phy with its 95% confidence interval for each category obtained from the linear regression models showed in S6 Table with the command marginsplot of Stata. https://doi.org/10.1371/journal.pone.0214717.g004 of swollen joints, at baseline that improved in terms of tender joints but neither in terms of swollen counts, nor APR or GDA-Pat. By contrast, the patient with no response by EULAR criteria but good response with HUPI was a female patient with extremely high number of ten- der joints that did not improve with treatment, whereas the remaining parameters were low at baseline and improved with treatment. PLOS ONE | https://doi.org/10.1371/journal.pone.0214717 April 10, 2019 9 / 19 Responsiveness of HUPI Table 3. Distribution of patients from ACT-RAY depending on clinical response assessed either by EULAR or HUPI criteria at different time-points. Week 12 (n = 518) Week 24 (n = 503) Week 52 (n = 412) EULAR HUPI HUPI HUPI None Moderate Good None Moderate Good None Moderate Good None 40 5 1 11 3 0 5 2 0 Moderate 70 164 42 34 99 36 19 54 23 Good 3 47 146 3 43 274 3 26 280 Same response 67.6% 76.3% 82.3% https://doi.org/10.1371/journal.pone.0214717.t003 Finally, to analyze which response set may be more accurate, we used as external criterion the ΔGDA-Phy from baseline to 12, 24 and 52 weeks. As shown in Fig 3C and S1 Fig panel B, in average an improvement of GDA-Phy�20 was associated with no response at all time points for HUPI response criteria and at week 12 for EULAR-RC. In the following time-points EULAR no response tended to be associated with lower ΔGDA-Phy. Regarding moderate and good responses, by HUPI criteria the average improvement in ΔGDA-Phy tended to be more lineal, whereas by EULAR criteria, higher improvement in ΔGDA-Phy were needed to reach moderate response and then lower improvements were needed to reach good response with respect moderate response (Fig 2C and 2D and S2 Fig panel B). This can be appreciated with the beta coefficients of the linear correlation models, in which the Akaike information criteria was always lower for the models run with HUPI than with EULAR response, suggesting that the former were better fitted (S6 Table). Similar results were observed in the PROAR and EMECAR cohorts (S3 Fig and S7 Table). Discussion HUPI was developed in an intent to provide a more accurate tool for assessing disease activity in patients with early RA and undifferentiated arthritis [12]. Validation is an ongoing process and new instruments like the HUPI need to be tested in different populations; therefore, we aimed to further validate HUPI by evaluating its responsiveness and the recently proposed HUPI-based response criteria [13]. This was particularly necessary in patients from clinical tri- als, whose baseline disease activity, as part of the general inclusion criteria, is usually very high. At present, there is no gold standard to assess disease activity in RA, nevertheless we used pooled indices of multiple measures that have been previously developed based on the Core Data Set proposed by Felson et al[24]. All in all, the data presented in this work showed that HUPI exhib- its comparable responsiveness to that of DAS28 and better than SDAI and CDAI. In addition, our data suggest that HUPI-based response criteria are slightly more stringent than EULAR’s. Baseline data from ACT-RAY have allowed confirming that in a clinical trial setting HUPI has a “ceiling effect”, likely due to its design. Remarkably, this ceiling effect was not detected in the other two cohorts more representative of patients seen in routine care. Nevertheless, in our opinion, patients with 5 or more swollen and tender joints and GDA-Pat higher than 50/100 and CRP levels higher 1 mg/dl show very high disease activity and need special therapeutic approaches irrespective of the magnitude of these variables above these limits. Despite this “ceiling effect”, HUPI showed the largest sensitivity to change in all three popu- lations, with SES superior to those of SDAI, CDAI and GDA, either by physician and patient. We recognize that SES is not the best statistic to report responsiveness, as it is only assessing internal responsiveness; however, it allows comparison across indices with varying range of values. In addition, similar results were reported using other methods when we described the index [12]. The poorer responsiveness of SDAI and CDAI may be a consequence of their design’s simplicity, leading to non-normally distributed variables with a highly spread range of PLOS ONE | https://doi.org/10.1371/journal.pone.0214717 April 10, 2019 10 / 19 Responsiveness of HUPI values in moderate and high disease activity. This can also explain the disparities in response measurements showed at Fig 2. On the other hand, since the responsiveness of HUPI is quite similar to that of DAS28, the response criteria based in both indices behave very similarly. Small differences have been detected, being HUPI slightly more stringent, with larger percentages of patients considered non-responders in ACT-RAY compared to percentages of EULAR response. These differences decreased along follow-up, although they were still detected at week 52, being the fast effect of tocilizumab on APR a possible explanation, since ESR is highly weighted in DAS28 [25]. Another possibility to explain this discrepancy is the tender joint count, that is also highly weighted in DAS28 and in HUPI is weighted differently by gender [12]. In this regard, it has been described that the presence of fibromyalgia can interfere with the assessment of disease activity with DAS28, since it impacts in the subjective components of the index, such as tender joint count[26]. Nevertheless, it is difficult to know whether being HUPI-based response more stringent than EULAR response may be an advantage or a disadvantage. In ACT-RAY, all patients were treated with tocilizumab plus placebo or methotrexate, showing no statistical differences between groups, but a statistical difference from baseline in both groups [15]. On these grounds, we considered patients from both groups experiencing a similar change; however, it was not our aim to evaluate treatment effect with any index. For this reason, we cannot deter- mine whether HUPI-based response is as stringent in a “real” placebo group as in an active treatment group, nor whether it helps discriminating the effect of the drug. Conclusion In summary, despite its “ceiling effect”, HUPI shows good responsiveness in all the scenarios tested. In addition, the response criteria based on this new index seems to be more stringent than the EULAR response criteria, although we need to deepen in the study of this characteris- tic to determine whether it could be more efficient to detect differences between placebo and active treatment. Supporting information S1 Fig. Comparison of EULAR and HUPI response criteria from baseline at different weeks of patients from ACT-RAY. A) Percentage of patients getting none, moderate or good response at week 24. B) Correlation of change in global disease assessment by physician (GDA-Phy) with the different categories of EULAR response and HUPI response at week 12 and week 24. Data in panels in section B are shown as the predicted mean change in GDA-Phy with its 95% confidence interval for each category obtained from the linear regression models showed in S6 Table. (TIF) S2 Fig. Comparison of EULAR and HUPI response criteria from baseline at different weeks of patients from PROAR and EMECAR. A) Percentage of patients getting none, mod- erate or good response at month 6, month 12 and month 24 in PROAR. B) Percentage of patients getting none, moderate or good response at 4 years of follow-up in EMECAR. (TIF) S3 Fig. Correlation of change in global disease assessment by physician (GDA-Phy) with the different categories of EULAR response and HUPI response in PROAR at months 6 (A), 12 (B) or 24 (C) and EMECAR at year 4th (D). Data are shown as the predicted mean change in GDA-Phy with its 95% confidence interval for each category obtained from the PLOS ONE | https://doi.org/10.1371/journal.pone.0214717 April 10, 2019 11 / 19 Responsiveness of HUPI linear regression models showed in S7 Table. (TIF) S1 Table. Scoring of the variables used to calculate HUPI. (DOCX) S2 Table. DMARD prescription during follow-up in PROAR. (DOCX) S3 Table. Number of visits from the different cohorts in which it was possible to calculate each index. (DOCX) S4 Table. Standardized size effects (95% confidence interval) of changes in disease activity assessed with HUPI and several other commonly used to assess response in rheumatoid arthritis. (DOCX) S5 Table. Characteristics of patients with opposite classification with EULAR and HUPI response criteria. (DOCX) S6 Table. Accuracy of EULAR-RC and HUPI-RC assessed by their correlation with change in GDA-Phy between baseline and different visits of ACT-RAY. (DOCX) S7 Table. Accuracy of EULAR and HUPI-based response criteria assessed by their correla- tion with ΔGDA-Phy in PROAR and ΔGDA-Pat in EMECAR. (DOCX) S1 File. EMECAR data in longitudinal format. This Excell file includes in longitudinal for- mat all data from EMECAR cohort used to develop the results shown in this paper. (XLS) S2 File. EMECAR data in wide format. This Excell file includes in wide format all data from EMECAR cohort used to develop the results shown in this paper. (XLS) S3 File. PROAR data in longitudinal format. This Excell file includes in longitudinal format all data from PROAR cohort used to develop the results shown in this paper. (XLS) S4 File. PROAR data in wide format. This Excell file includes in wide format all data from PROAR cohort used to develop the results shown in this paper. (XLS) Acknowledgments We want to thank all researchers involved in data collection at: ACT-RAY Lead authors: Dougados, M. CHU Paris Centre—Hôpital Cochin. France. E-mail: maxime. dougados@cch.ap-hop-paris.fr; Huizinga, T. Leiden University Medical Center, Netherlands. E-mail: T.W.J.Huizinga@lumc.nl PLOS ONE | https://doi.org/10.1371/journal.pone.0214717 April 10, 2019 12 / 19 Responsiveness of HUPI Abu Shakra, M. Soroka Medical Center. Israel Alberts, A. West Broward Rheumatology Associates, Inc. United States Alperi Lopez, M. Hospital Univ. Central de Asturias. Spain Amital, H. Chaim Sheba Medical Center. Israel Aringer, M. Universitätsklinikum "Carl Gustav Carus". Germany Aslanidis, S. Hippokratio Hospital. Greece Berenbaum, F. Hopital Saint Antoine. France Bijlsma, H. Academisch Medisch Centrum Utrecht. Netherlands Blanco Garcia, FJ. Complejo Hospitalario Universitario A Coruña. Spain Bliddal, H. Frederiksberg Sygehus. Denmark Borofsky, M. Clinical Research Center of Reading. United States Brocq, O. Ch Princesse Grace. Monaco Buldakov, S. Republican Clinicodiagnostic Center. Russian Federation Cantini, F. Presidio Ospedaliero Misericordia e Dolce. Italy Carreño Perez, L. Hospital General Universitario Gregorio Marañon. Spain Chahade, W. Hospital Estadual do Servidor Publico. Brazil Ciconelli, R. Universidade Federal de Sao Paulo. Brazil Codreanu, C. Centrul de Boli Reumatismale Dr. Ioan Stoia. Romania Dahlqvist, SR. Norrlands Universitary Hospital. Sweden Damjanov, N. Institut Za Reumatologiju. Serbia Diamantopoulos, A. Sørlandet Sykehus Kristiansand. Norway Dimdina, L. Clinical University Hospital Gailezers. Latvia Dimic, A. Institut Za Prevenciju, Lecenje I Rehabilitaciju. Serbia Dorokhov, A. State Institution of Health Care—Territorial Clinical Hospital. Russian Federation Dubikov, A. City Clinical Hospital # 2. Russian Federation Fadienko, G. Glpu Tjumen Regional Clinical Hospital #1. Russian Federation Fanø, N. Sjællands Universitetshospital, Køge. Denmark Ferreira, G. Hospital das Clinicas–UFMG. Brazil Gabrielli, A. Uni Politecnica Delle Marche; Ist. Di Clinica Medica Generale Ematologia Ed Immunologia Clinica. Italy Gaffney, K. Norfolk & Norwich Hospital. United Kingdom Gaudin, P. Hopital Sud. France Gerlag, DM. Academisch Medisch Centrum. Netherlands Gerli, R. Osp S. Maria Misericordia Dip. Italy Gonçalves, CR. Hospital das Clinicas–FMUSP. Brazil Hansen, MS. Gentofte Hospital. Denmark Hanvivadhanakul, P. Thammasat University Hospital. Thailand Høili, C. Sykehuset Ostfold Moss HF. Norway Hou, A. Inland Rheumatology; Clinical Trials, Inc. United States Hunter, J. Gartnavel General Hospital. United Kingdom Ilic, T. Clinical Centre of Vojvodina. Serbia Ionescu, R. Spitalul Sf Maria. Romania Kaine, J. Sarasota Arthritis Center. United States Kakurina, N. Clinical Hospital of Daugavpils. Latvia Kamalova, R. Republican clinical hospital. Russian Federation Kelly, T. Innovative Health Research. United States Knyazeva, L. GMU Kursk Regional Clinical Hospital. Russian Federation Krumina, L. L.Krumina GP practice. Latvia PLOS ONE | https://doi.org/10.1371/journal.pone.0214717 April 10, 2019 13 / 19 Responsiveness of HUPI Kurthen, R. Praxis Dr. med. Reiner Kurthen. Germany Lagrone, RP. St. Thomas Hospital. United States Lapadula, G. Ospedale Policlinico Di.M.I.M.P. Italy Lavrentjevs, V. P.Stradins Clinical University Hospital. Latvia Lawson, JG. Piedmont Arthritis Clinic. United States Lazic, Z. Clinical Center Kragujevac. Serbia Lejnieks, A. Rakus Clinic Linezers. Latvia Levy, Y. Meir Medical Center. Israel Lexberg, Å. Drammen sykehus Vestre Viken HF. Norway Mader, R. Haemek Hospital. Israel Mariette, X. Ch De Bicêtre. France Markovits, D. Rambam Medical Center. Israel Martin Mola, E. Htal. La Paz. Spain Maugars, Y. Hopital Hotel Dieu Et Hme. France Maymo Guarch, J. Hospital del Mar. Spain Mazurov, VI. Sbei Of Hpe "Northwestern State Medical University N.A. I.I.Mechnikov". Russian Federation Mikkelsen, K. Revmatismesykehuset. Norway Morovic Vergles, J. Clinical Hospital Dubrava. Croatia Nabizadeh, S. Martina Hansen Hospital. Norway Nanagara, R. Khon Kaen University. Thailand Nasonov, EL. Fsbi "Scientific Research Institute Of Rheumatology" Of Russian Academy Of Medical Sciences. Russian Federation. Navarro Sarabia, F. Hospital Universitario Virgen Macarena. Spain Neumann, T. Universitätsklinikum Jena. Germany Novak, S. Rheumatology and Clinical Immunology. Croatia Olech, E. Oklahoma Medical Research Foundation. United States Oza, M. Arthritis/Osteoporosis Treatment Center. United States Paran, D. Sourasky / Ichilov Hospital. Israel Parsik, E. North Estonian Regional Hospital. Estonia Pegram, S. Rheumatic Disease Clin Res Ctr. United States Pombo Suarez, M. Hospital Nuestra Señora de la Esperanza. Spain Popova, T. Municipal Autonomous Institution Of Healthcare "City Clinical Hospital #40". Russian Federation Puechal, X. Ch Du Mans. France Raja, N. Agilence Arthritis and Osteoporosis Medical Center, Inc. United States Ridley, D. St. Paul Rheumatology. United States Rosner, I. Bnei Zion Medical Center. Israel Rubbert-Roth, A. Klinik der Uni zu Köln. Germany Rudin, A. Sahlgrenska Universitetssjukhuset. Sweden Saraux, A. Hopital La Cavale Blanche. France Saulite-Kandevica, D. D.Saulite-Kandevica Private Practice. Latvia Settas, L. Ahepa Hospital. Greece Sfikakis, P. Laiko General Hospital. Greece Sheeran, T. Cannock Chase Hospital. United Kingdom Sizikov, A. FSBI Scientific Research Institute of Clinical Immunology of SB of RAMS. Rus- sian Federation Stamenkovic, D. Clinical Hospital Centre Rijeka. Croatia Stefanovic, D. Military Medical Academy. Serbia PLOS ONE | https://doi.org/10.1371/journal.pone.0214717 April 10, 2019 14 / 19 Responsiveness of HUPI Stolow, JB. Texas Arthritis Research Center. United States Tan, AL. Chapel Allerton Hospital. United Kingdom Tebib, J. Ch Lyon Sud. France Tishler, M. Assaf Harofe. Israel Tony, HP. Universitätsklinikum Würzburg. Germany Troum, OM. United States Uaratanawong, S. Vajira Hospital. Thailand Ucar Angulo, E. Hospital de Basurto. Spain Valenzuela, G. Berma Research Group. United States van der Laken, K. VU Medisch Centrum. Nehterlands Van Laar, J. School of Clinical Medical Services. United Kingdom van RIEL, P.L.C.M. Akademisch Ziekenhuis St. Radboud. Netherlands Vasilopoulos, D. Hippocrateio Hospital of Athens. Greece Veldi, T. East Tallinn Central Hospital. Estonia Vinogradova, I. State Institution of Healthcare Ulyanovsk Regional Clinical Hospital. Rus- sian Federation Vosse, D. Academisch Ziekenhuis Maastricht. Netherlands Wassenberg, S. Evangelisches Fachkrankenhaus. Germany Weidmann, C. Medvin Clinical Research. United States Weitz, M. Center For Arthritis. United States Wollenhaupt, J. Schön Klinik Hamburg-Eilbek Klinik für Rheumatologie. Germany Xavier, R. Hospital das Clinicas–UFRGS. Brazil Yakupova, S. Kazan State Medical University. Russian Federation Zagar, I. Klinicki Bolnicki Centar Zagreb. Croatia Zavgorodnaja, T. P.Stradins Clinical University Hospital. Latvia Zemerova, E. Khanty-Mansiysk Autonomous Area—Ugri Region Clinical Hospital. Rus- sian Federation Zisman, D. Carmel Hospital. Israel Zonova, E. FSBI Scientific Research Institute of Clinical and Experimental Lymphology of SB of RAMS. Russian Federation EMECAR Lead author: Loreto Camona, InMusc. Spain. E-mail: loreto.carmona@inmusc.eu Abasolo Alcazar L, Hospital Clı́nico Universitario San Carlos, Madrid. Spain Alegre de Miguel C, Hospital de Malalties Reumatiques, Barcelona. Spain Andreu Sánchez JL, Clı́nica Puerta de Hierro. Majadahonda. Spain Aragón Dı́ez A, Hospital Nuestra Señora del Prado. Talavera de la Reina. Spain Balsa Criado A, Hospital La Paz. Madrid. Spain Batlle Gualda E, Hospital General Universitario de Alicante. Spain Belmonte Serrano MA, Hospital General de Castellón. Spain Beltrán Audera J, Hospital Clı́nico Universitario de Zaragoza. Spain Beltrán Fabregat J, Hospital General de Castellón. Spain Bonilla Hernan G, Hospital La Paz. Madrid. Spain Caro Fernández N, Hospital Nuestra Señora del Prado. Talavera de la Reina. Spain Casado E, Hospital Universitario Germans Trias i Pujol. Badalona. Spain Cebrian Mendez L, Hospital Gregorio Marañón. Madrid. Spain Corteguera Coro M, Hospital Nuestra Señora de Sonsoles. Avila. Spain Cuadra Dı́az JL, Hospital Nuestra Señora del Carmen. Ciudad Real. Spain PLOS ONE | https://doi.org/10.1371/journal.pone.0214717 April 10, 2019 15 / 19 Responsiveness of HUPI Cuesta E, Hospital Virgen de La Luz. Cuenca. Spain Fiter Aresté J, Hospital Son Dureta. Palma de Mallorca. Spain Freire Gonzalez M, Hospital Gregorio Marañón. Madrid. Spain Galindo Izquierdo M, Hospital 12 de Octubre. Madrid. Spain Garcı́a Meijide JA, Hospital Clı́nico Universitario de Santiago de Compostela. Spain Garcı́a Gómez MC, Hospital Universitario de Bellvitge. Hospitalet de Llobregat. Spain Giménez Ubeda E, Hospital Clı́nico Universitario de Zaragoza. Spain Gómez Centeno E, Hospital Clinic i Provincial. Barcelona. Spain Gómez Vaquero C, Hospital Universitario de Bellvitge. Hospitalet de Llobregat. Spain Gon- zález Fernández MJ, Hospital de Malalties Reumatiques. Barcelona. Spain González Gómez ML, Hospital Gregorio Marañón. Madrid. Spain González Hernández T, Instituto Provincial de Rehabilitación. Madrid. Spain González-Alvaro I, Hospital Universitario la Princesa. Madrid. Spain González-Montagut Gómez C, Hospital Virgen de La Luz. Cuenca. Spain Grandal Delgado Y, Hospital General de Jerez de La Frontera. Spain Gratacos Masmitja J, Complejo Hospitalario del Parc Tauli. Sabadell. Spain Hernández del Rı́o A, Hospital Juan Canalejo. A Coruña. Spain Instxaurbe AR, Hospital de Basurto. Bilbao. Spain Irigoyen Oyarzabal MV, Hospital General Carlos Haya. Malaga. Spain Jiménez Palop M, Hospital Nuestra Señora de Sonsoles. Avila. Spain Juan Mas A, Hospital Son Llatzer. Palma de Mallorca. Spain Júdez Navarro E, Hospital Clı́nico Universitario San Carlos. Madrid. Spain Larrosa Padro M, Complejo Hospitalario del Parc Tauli. Sabadell. Spain López Longo FJ, Hospital Gregorio Marañón. Madrid. Spain Loza Santamaria E, Hospital Clı́nico Universitario San Carlos. Madrid. Spain Maese Manzano J, Fundación Española de Reumatologı́a. Manero Ruiz FJ, Hospital Clı́nico Universitario de Zaragoza. Spain Mateo Bernardo I, Hospital 12 de Octubre. Madrid. Spain Mayordomo González L, Hospital Universitario de Valme. Sevilla. Spain Mazzucheli R, Hospital Fundación de Alcorcón. Spain Medrano San Idelfonso M, Hospital Clı́nico Universitario de Zaragoza. Spain Naranjo Hernández A, Hospital de Gran Canaria Dr. Negrı́n. Las Palmas. Spain Pecondón Español A, Hospital Clı́nico Universitario de Zaragoza. Spain Peiró Callizo E, Hospital Virgen de La Luz. Cuenca. Spain Quirós Donate J, Hospital Fundación de Alcorcón. Spain Ramos López P, Hospital Prı́ncipe de Asturias. Alcala de Henares. Spain Rivera Redondo J, Instituto Provincial de Rehabilitación. Madrid. Spain Rodrı́guez Gómez M, Complejo Hospitalario Cristal-Piñor. Ourense. Spain Rodrı́guez López M, Hospital Arquitecto Marcide. Lugo. Spain Roselló Pardo R, Hospital General San Jorge. Huesca. Spain Sampedro Alvarez J, Hospital Virgen de La Salud. Toledo. Spain Sanmartı́ Sala R, Hospital Clinic i Provincial. Barcelona. Spain Santos Rey Rey J, Hospital Virgen de La Salud. Toledo. Spain Tena Marsá X, Hospital Universitario Germans Trias i Pujol. Badalona. Spain Tenorio Martı́n M, Hospital del Insalud-Ceuta. Spain Torres Martı́n MC, Hospital Nuestra Señora de Sonsoles. Avila. Spain Ureña Garnica I, Hospital General Carlos Haya. Malaga. Spain Valdazo de Diego JP, Hospital General Virgen de La Concha. Zamora. Spain Valls M, Hospital Universitario Germans Trias i Pujol. Badalona. Spain PLOS ONE | https://doi.org/10.1371/journal.pone.0214717 April 10, 2019 16 / 19 Responsiveness of HUPI Villaverde Garcı́a V, Hospital La Paz. Madrid. Spain Zarco Montejo P, Hospital Fundación de Alcorcón. Spain Zubieta Tabernero J, Hospital Virgen de La Salud. Toledo. Spain PROAR Lead authors: Alejandro Balsa. Hospital Universitario La Paz, Madrid. Spain. E-mail: alejan- dro.balsa@salud.madrid.org; Raimon Sanmartı́. Hospital Clinic, Barcelona. Spain. E-mail: san- marti@clinic.ub.es Cabezas, JA. Complejo Hospitalario San Millan-San Pedro, Logroño. Spain Cantalejo, M. Hospital Univ de Getafe. Spain Chamizo, E. Hospital de Mérida. Spain Ciruelo, E. Hospital Gral de Segovia. Spain Corrales, A. Hospital Comarcal de Laredo. Spain Cruz, A. Hospital Severo Ochoa, Leganés. Spain Dı́az, C. Hospital de La Santa Creu i Sant Pau, Barcelona. Spain Fiter, J. H Son Dureta, Palma de Mallorca. Spain Freire, MM. Hospital Juan Canalejo, A Coruña. Spain Galindo, M. Hospital 12 de Octubre, Madrid. Spain Garcı́a de Vicuña, MR. Hospital La Princesa, Madrid. Spain Gelman, SM. Hospital Gral de Manresa, Spain González Crespo, R. Hospital 12 de Octubre, Madrid. Spain González Fernández, C. Hospital Gregorio Marañon, Madrid. Spain González Fernández, MJ. Institut Dexeus, Barcelona. Spain González Hernández, T. Instituto Prov de Rehabilitación, Madrid. Spain Gracia, A. Hospital de Sagunto. Spain Granados, J. Hospital Mutua de Terrassa. Spain Guzman, MA. Hospital Virgen de las Nieves, Granada. Spain Irigoyen, MV. Hospital Gral Carlos Haya, Malaga. Spain Juan, A. H Son Dureta, Palma de Mallorca. Spain Juanola, X. (Hospital de Bellvitge, L’Hospitalet de Llobregat. Spain Laiz, A. Hospital de La Santa Creu i Sant Pau, Barcelona. Spain Manero, FJ. Hospital Univ Miguel Servet, Zaragoza. Spain Martı́nez, A. Hospital Gral Univ de Alicante. Spain Martinez, F. Hospital Univ Reina Sofı́a, Córdoba. Spain Mata, C. Hospital Comarcal Sierrallana, Torrelavega. Spain Maymo, J. Imas, Barcelona. Spain Navarro, FJ. Hospital Gral Univ de Elche. Spain Peiro, E. Hospital Virgen de La Luz, Cuenca. Spain Pérez, F. Hospital Gral de Requena. Spain Pérez, G. Hospital del Insalud, Ceuta. Spain Pérez, M. Hospital Gral Carlos Haya, Malaga. Spain Pujol, M. Hospital Mutua de Terrassa. Spain Quirós, J. Hospital Fundación Alcorcón. Spain Ribas, B. Hospital San Juan de Dios, Palma de Mallorca. Spain Riera, M. Hospital Creu Roja, Barcelona. Spain Rivera, J. Instituto Prov de Rehabilitación, Madrid. Spain Rodrı́guez, JM. Hospital Univ de Getafe. Spain Rodrı́guez Gómez, M. Complejo Hospitalario Cristal-Piñor, Orense. Spain PLOS ONE | https://doi.org/10.1371/journal.pone.0214717 April 10, 2019 17 / 19 Responsiveness of HUPI Roselló, R. Hospital Gral San Jorge, Huesca. Spain Tenorio, M. Hospital del Insalud, Ceuta. Spain Toyos, FJ. Hospital Univ Virgen Macarena, Sevilla. Spain Author Contributions Conceptualization: Isidoro González-Álvaro, Isabel Castrejón, Loreto Carmona. Data curation: Isidoro González-Álvaro. Formal analysis: Isidoro González-Álvaro. Funding acquisition: Isidoro González-Álvaro. Methodology: Isabel Castrejón, Loreto Carmona. Supervision: Loreto Carmona. Writing – original draft: Isidoro González-Álvaro. Writing – review & editing: Isidoro González-Álvaro, Isabel Castrejón, Loreto Carmona. References 1. Anderson J, Caplan L, Yazdany J, Robbins ML, Neogi T, Michaud K, et al. Rheumatoid arthritis disease activity measures: American College of Rheumatology recommendations for use in clinical practice. Arthritis Care Res (Hoboken). 2012; 64(5):640–7. Epub 2012/04/05. https://doi.org/10.1002/acr.21649 PMID: 22473918. 2. Saag KG. Low-dose corticosteroid therapy in rheumatoid arthritis: balancing the evidence. The Ameri- can journal of medicine. 1997; 103(6A):31S–9S. PMID: 9455967 3. Singh JA, Saag KG, Bridges SL Jr., Akl EA, Bannuru RR, Sullivan MC, et al. 2015 American College of Rheumatology Guideline for the Treatment of Rheumatoid Arthritis. Arthritis Care Res (Hoboken). 2016; 68(1):1–25. Epub 2015/11/08. https://doi.org/10.1002/acr.22783 PMID: 26545825. 4. Smolen JS, Landewe R, Bijlsma J, Burmester G, Chatzidionysiou K, Dougados M, et al. EULAR recom- mendations for the management of rheumatoid arthritis with synthetic and biological disease-modifying antirheumatic drugs: 2016 update. Ann Rheum Dis. 2017. Epub 2017/03/08. https://doi.org/10.1136/ annrheumdis-2016-210715 PMID: 28264816. 5. Aletaha D, Landewe R, Karonitsch T, Bathon J, Boers M, Bombardier C, et al. Reporting disease activity in clinical trials of patients with rheumatoid arthritis: EULAR/ACR collaborative recommendations. Ann Rheum Dis. 2008; 67(10):1360–4. https://doi.org/10.1136/ard.2008.091454 PMID: 18791055. 6. Leeb BF, Haindl PM, Maktari A, Nothnagl T, Rintelen B. Disease activity score-28 values differ consid- erably depending on patient’s pain perception and sex. J Rheumatol. 2007; 34(12):2382–7. PMID: 17985407. 7. Sokka T, Toloza S, Cutolo M, Kautiainen H, Makinen H, Gogus F, et al. Women, men, and rheumatoid arthritis: analyses of disease activity, disease characteristics, and treatments in the QUEST-RA Study. Arthritis Res Ther. 2009; 11(1):R7. https://doi.org/10.1186/ar2591 PMID: 19144159. 8. Ahlmen M, Svensson B, Albertsson K, Forslind K, Hafstrom I. Influence of gender on assessments of disease activity and function in early rheumatoid arthritis in relation to radiographic joint damage. Ann Rheum Dis. 2010; 69(1):230–3. https://doi.org/10.1136/ard.2008.102244 PMID: 19158113. 9. Castrejón Fernández I, Martı́nez-López J, Ortiz Garcı́a A, Carmona Ortells L, Garcı́a-Vicuña R, Gonza- lez-Alvaro I. Influencia del género en la respuesta al tratamiento en una cohorte de pacientes con artritis reumatoide precoz del área 2 de la Comunidad de Madrid. Reumatol Clin. 2010; 6(3):134–40. 10. Barnabe C, Bessette L, Flanagan C, Leclercq S, Steiman A, Kalache F, et al. Sex differences in pain scores and localization in inflammatory arthritis: a systematic review and metaanalysis. J Rheumatol. 2012; 39(6):1221–30. Epub 2012/04/17. https://doi.org/10.3899/jrheum.111393 PMID: 22505697. 11. Couderc M, Gottenberg JE, Mariette X, Pereira B, Bardin T, Cantagrel A, et al. Influence of gender on response to rituximab in patients with rheumatoid arthritis: results from the Autoimmunity and Rituximab registry. Rheumatology (Oxford). 2014; 53(10):1788–93. Epub 2014/05/14. https://doi.org/10.1093/ rheumatology/keu176 PMID: 24821852. 12. Castrejon I, Carmona L, Ortiz AM, Belmonte MA, Martinez-Lopez JA, Gonzalez-Alvaro I. Development and validation of a new disease activity index as a numerical sum of four variables in patients with early PLOS ONE | https://doi.org/10.1371/journal.pone.0214717 April 10, 2019 18 / 19 Responsiveness of HUPI arthritis. Arthritis Care Res (Hoboken). 2013; 65(4):518–25. Epub 2012/09/25. https://doi.org/10.1002/ acr.21854 PMID: 23002022. 13. Gonzalez-Alvaro I, Castrejon I, Ortiz AM, Toledano E, Castaneda S, Garcia-Vadillo A, et al. Cut-Offs and Response Criteria for the Hospital Universitario La Princesa Index (HUPI) and Their Comparison to Widely-Used Indices of Disease Activity in Rheumatoid Arthritis. PLoS One. 2016; 11(9):e0161727. Epub 2016/09/08. https://doi.org/10.1371/journal.pone.0161727 PMID: 27603313. 14. Arnett FC, Edworthy SM, Bloch DA, McShane DJ, Fries JF, Cooper NS, et al. The American Rheuma- tism Association 1987 revised criteria for the classification of rheumatoid arthritis. Arthritis Rheum. 1988; 31(3):315–24. PMID: 3358796 15. Dougados M, Kissel K, Sheeran T, Tak PP, Conaghan PG, Mola EM, et al. Adding tocilizumab or switching to tocilizumab monotherapy in methotrexate inadequate responders: 24-week symptomatic and structural results of a 2-year randomised controlled strategy trial in rheumatoid arthritis (ACT-RAY). Ann Rheum Dis. 2013; 72(1):43–50. Epub 2012/05/09. https://doi.org/10.1136/annrheumdis-2011- 201282 PMID: 22562983. 16. Huizinga TW, Conaghan PG, Martin-Mola E, Schett G, Amital H, Xavier RM, et al. Clinical and radio- graphic outcomes at 2 years and the effect of tocilizumab discontinuation following sustained remission in the second and third year of the ACT-RAY study. Ann Rheum Dis. 2015; 74(1):35–43. Epub 2014/08/ 30. https://doi.org/10.1136/annrheumdis-2014-205752 PMID: 25169728. 17. Villaverde Garcia V, Balsa A, Carmona L, Sanmarti R, Maese J, Pascual D, et al. [What are patients with early rheumatoid arthritis like in Spain? Description of the PROAR cohort]. Reumatol Clin. 2009; 5(3):115–20. Epub 2009/05/01. PMID: 21794592. 18. Gonzalez-Alvaro I, Descalzo MA, Carmona L. Trends towards an improved disease state in rheumatoid arthritis over time: influence of new therapies and changes in management approach: analysis of the EMECAR cohort. Arthritis Res Ther. 2008; 10(6):R138. https://doi.org/10.1186/ar2561 PMID: 19036152. 19. Husted JA, Cook RJ, Farewell VT, Gladman DD. Methods for assessing responsiveness: a critical review and recommendations. J Clin Epidemiol. 2000; 53(5):459–68. PMID: 10812317. 20. Durlak JA. How to select, calculate, and interpret effect sizes. Journal of pediatric psychology. 2009; 34 (9):917–28. Epub 2009/02/19. https://doi.org/10.1093/jpepsy/jsp004 PMID: 19223279. 21. Hedges L, Olkin I. Statistical Methods for Meta-Analysis. New York: Academic Press; 1985. 22. Fransen J, van Riel PL. The Disease Activity Score and the EULAR response criteria. Clin Exp Rheu- matol. 2005; 23(5 Suppl 39):S93–9. Epub 2005/11/09. PMID: 16273792. 23. Akaike H. A new look at the statistical model identification. IEEE transactions on Automatic Control. 1974; 19(6):716–23. 24. Felson DT, Anderson JJ, Boers M, Bombardier C, Chernoff M, Fried B, et al. The American College of Rheumatology preliminary core set of disease activity measures for rheumatoid arthritis clinical trials. The Committee on Outcome Measures in Rheumatoid Arthritis Clinical Trials. Arthritis Rheum. 1993; 36 (6):729–40. Epub 1993/06/01. PMID: 8507213. 25. Prevoo ML, van ’t Hof MA, Kuper HH, van Leeuwen MA, van de Putte LB, van Riel PL. Modified disease activity scores that include twenty-eight-joint counts. Development and validation in a prospective longi- tudinal study of patients with rheumatoid arthritis. Arthritis Rheum. 1995; 38(1):44–8. PMID: 7818570. 26. Salaffi F, Di Carlo M, Carotti M, Sarzi-Puttini P. The subjective components of the Disease Activity Score 28-joints (DAS28) in rheumatoid arthritis patients and coexisting fibromyalgia. Rheumatol Int. 2018; 38(10):1911–8. Epub 2018/06/30. https://doi.org/10.1007/s00296-018-4096-z PMID: 29955927. PLOS ONE | https://doi.org/10.1371/journal.pone.0214717 April 10, 2019 19 / 19