Browse the corpus

Introduction Rituximab, a chimeric anti-CD20 monoclonal antibody (mAb) remains an important treatment option for moderate to severe systemic lupus erythematosus (SLE). A high degree of efficacy of rituximab across a range of lupus manifestations has been reported in open-label studies from single-centre series,1–3 multicentre registries4–6 and a systematic review of off-label use.7 Despite the success of these series, two phase III randomised placebo-controlled trials in non-renal lupus8 and renal lupus9 failed to meet their primary end-points. The discrepancy between the randomised trials and real-world evidence has been attributed to aspects of trial design including choice of end-points, the use of an active comparator, inclusion criteria and low statistical power.10 Nevertheless, there are also mechanistic reasons for the failure of rituximab in clinical trials in SLE. B-cell killing by rituximab appeared less efficient in SLE than rheumatoid arthritis (RA)11 due to internalisation through interaction with FcγRIIb resulting in reduced effector activity12 and pathogenic lupus autoantibodies that were produced by long-lived plasma cells.13 14 Using highly sensitive flow cytometry (HSFC), a protocol that was optimised for the detection of plasmablasts, we discovered that the depth of B-cell depletion predicted response in RA15 and SLE.2 Similar studies as well as identifying other clinical predictors of response to rituximab in SLE are needed to optimise its use and to help design trials of alternative B-cell depleting strategies.

d for the detection of plasmablasts, we discovered that the depth of B-cell depletion predicted response in RA15 and SLE.2 Similar studies as well as identifying other clinical predictors of response to rituximab in SLE are needed to optimise its use and to help design trials of alternative B-cell depleting strategies. B-cell depletion therapy with rituximab is transient. Some patients with initial good response experience relapse after B-cell repopulation (although with a variable interval). In our published discovery cohort, we showed a bimodal pattern of relapse. Earlier relapse requiring rituximab retreatment was predicted by a plasmablast count of >0.0008×109/L at 6 months (the time of initial clinical response).2 Patients with lower plasmablasts at 6 months had sustained response without retreatment. Validation of this as a biomarker is therefore needed to determine whether HSFC can be used in clinical practice to guide retreatment decisions.

asmablast count of >0.0008×109/L at 6 months (the time of initial clinical response).2 Patients with lower plasmablasts at 6 months had sustained response without retreatment. Validation of this as a biomarker is therefore needed to determine whether HSFC can be used in clinical practice to guide retreatment decisions. Repeat treatment with rituximab is effective.1 However, we observed cases of patients with SLE who had previously depleted and responded well to rituximab but subsequently developed (1) a severe infusion reaction >24 hours during the second infusion of a cycle, (2) failure to deplete CD20+ (naïve and memory) B-cells and (3) clinical non-response during repeat cycles. We called this phenomenon secondary non-depletion and non-response (2NDNR), which was suggestive of immunogenicity to rituximab and could be overcome by alternative anti-CD20 mAbs, particularly humanised. Therefore, the aims of the study were to assess factors predicting primary and secondary non-response to rituximab in SLE including validation of B-cell depletion and to evaluate management of 2NDNR using alternative anti-CD20 agents. Methods Patients and design A prospective observational study was conducted of all patients with moderate to severe SLE who were treated with rituximab in Leeds between January 2004 and July 2016. Inclusion criteria included (1) adults (>16 years old); (2) fulfilling the revised 1997 American College of Rheumatology classification for SLE16 and (3) at least 6 months follow-up post-rituximab.

d of all patients with moderate to severe SLE who were treated with rituximab in Leeds between January 2004 and July 2016. Inclusion criteria included (1) adults (>16 years old); (2) fulfilling the revised 1997 American College of Rheumatology classification for SLE16 and (3) at least 6 months follow-up post-rituximab. Treatment protocol All patients received a first cycle of therapy consisting of 100 mg of methylprednisolone and 1000 mg of rituximab given intravenously on days 1 and 14. Further cycles of the same regimen were repeated on clinical relapse (defined below). Of those who met 2NDNR criteria, their treatment was switched from rituximab to humanised anti-CD20 mAbs either by using (1) 2×1000 mg ocrelizumab (compassionate use from Roche UK) or (2) 2×700 mg ofatumumab (individual funding request to NHS England).

Treatment protocol All patients received a first cycle of therapy consisting of 100 mg of methylprednisolone and 1000 mg of rituximab given intravenously on days 1 and 14. Further cycles of the same regimen were repeated on clinical relapse (defined below). Of those who met 2NDNR criteria, their treatment was switched from rituximab to humanised anti-CD20 mAbs either by using (1) 2×1000 mg ocrelizumab (compassionate use from Roche UK) or (2) 2×700 mg ofatumumab (individual funding request to NHS England). Clinical data and outcomes Disease activity was assessed using the British Isles Lupus Assessment Group (BILAG-2004)17 at baseline and every 3 months thereafter. Clinical responses at 6 months were determined as following: (1) major clinical response=improvement of all domains rated A/B to grade C/better and no A/B flare between baseline and 6 months; (2) partial clinical response=maximum of 1 domain with a persistent grade B with improvement in all other domains and no A or B flare and (3) non-response=those not meeting the criteria for major or partial clinical response. Relapse was defined as a new grade A or recurrence of ≥1 grade B following either major/partial clinical response at 6 months. Global BILAG score was calculated as follows: grade A=12, grade B=8, grade C=1 and grades D and E=0.18

nd (3) non-response=those not meeting the criteria for major or partial clinical response. Relapse was defined as a new grade A or recurrence of ≥1 grade B following either major/partial clinical response at 6 months. Global BILAG score was calculated as follows: grade A=12, grade B=8, grade C=1 and grades D and E=0.18 Laboratory assessments Peripheral blood B-cell subsets (naïve, memory B-cells and plasmablasts) were measured using HSFC as previously described15 at baseline, 6 months and every 6 months without knowledge of clinical status other than time since rituximab. Complete B-cell depletion was defined as counts <0.0001×109/L and repopulation as ≥0.0001×109/L. Anti-dsDNA antibody titres were measured by ELISA until July 2012 and Bioplex 2200 Immunoassay (after July 2012). Complement levels (C3 and C4) and total serum immunoglobulin titres were measured by nephelometry. Anti-rituximab antibodies were tested on a subset of patients with 2NDNR using the Promonitor® Anti-Rituximab ELISA according to the manufacturer’s instructions and compared these concentrations to those with continued response to rituximab. A positive test (as determined by the manufacturer) was concentration >140 AU/mL. Safety Serious infections were recorded irrespective of suspected association with SLE and/or therapy. These were infections that resulted in hospitalisation for >24 hours or required intravenous antibiotics. Details about other safety assessment can be found in online supplementary files.

Anti-rituximab antibodies were tested on a subset of patients with 2NDNR using the Promonitor® Anti-Rituximab ELISA according to the manufacturer’s instructions and compared these concentrations to those with continued response to rituximab. A positive test (as determined by the manufacturer) was concentration >140 AU/mL. Safety Serious infections were recorded irrespective of suspected association with SLE and/or therapy. These were infections that resulted in hospitalisation for >24 hours or required intravenous antibiotics. Details about other safety assessment can be found in online supplementary files. Statistical analysis Descriptive statistics were summarised using mean with SD or median with IQR for continuous variables and proportion for categorical variables. Multiple imputation was used for missing data. Multivariable analyses were performed using logistic regression after checking for multicollinearity. The significance of the association between categorical variables was tested by Fisher’s exact test, while for continuous variables using Mann-Whitney U test. Receiver operator curves (ROCs) were used to measure sensitivity and specificity of optimal thresholds for investigations predicting time-to-clinical relapse. All statistical analysis was performed using Stata V.13.1 and Graph Pad Prism V.6.01 for Windows.

Statistical analysis Descriptive statistics were summarised using mean with SD or median with IQR for continuous variables and proportion for categorical variables. Multiple imputation was used for missing data. Multivariable analyses were performed using logistic regression after checking for multicollinearity. The significance of the association between categorical variables was tested by Fisher’s exact test, while for continuous variables using Mann-Whitney U test. Receiver operator curves (ROCs) were used to measure sensitivity and specificity of optimal thresholds for investigations predicting time-to-clinical relapse. All statistical analysis was performed using Stata V.13.1 and Graph Pad Prism V.6.01 for Windows. Results Patient characteristics Of 125 patients with SLE who were treated with rituximab at our unit, 117 patients with evaluable data at 6 months were studied. Baseline characteristics are described in table 1. One hundred and twelve (96%) had refractory and active disease as defined by BILAG ≥1A score and/or ≥2B scores. The remaining five had BILAG B in one domain only but was refractory to other conventional therapies as well as on maintenance with oral prednisolone ≥10 mg daily. Total follow-up was 492 patient-years. Table 1 Baseline characteristics of the 117 patients with SLE treated with rituximab

Results Patient characteristics Of 125 patients with SLE who were treated with rituximab at our unit, 117 patients with evaluable data at 6 months were studied. Baseline characteristics are described in table 1. One hundred and twelve (96%) had refractory and active disease as defined by BILAG ≥1A score and/or ≥2B scores. The remaining five had BILAG B in one domain only but was refractory to other conventional therapies as well as on maintenance with oral prednisolone ≥10 mg daily. Total follow-up was 492 patient-years. Table 1 Baseline characteristics of the 117 patients with SLE treated with rituximab Age at first RTX infusion, median (IQR) years 39 (26–52) No. female patient (%) 109 (93) Ethnicity, N (%) Caucasian 80 (68) Afro-Caribbean 11 (10) South Asian 20 (17) Other 6 (5) SLE disease duration at first RTX, median (IQR) years 6 (2–11) Positive ANA at diagnosis, N (%) 117 (100) Antibody status at first RTX infusion, N (%) Positive 108 (92) anti-dsDNA 56 (48) Anti-Ro 57 (49) Anti-La 18 (15) Anti-Smith 15 (13) Anti-Chromatin 19 (16) Anti-RNP 23 (20) Anti-Ribosomal P 6 (5) Anti-Cardiolipin/anti-B2-glycoprotein 14 (12) Prior CYC therapy, N (%) 63 (54) Cumulative dose of CYC, mean ± SD gram   6.6 ± 4.2 Number of prior immunosuppressant failure (including CYC but excluding glucocorticoid), median (range) 3 (0–9) Concomitant antimalarials, N (%) 88 (75) Concomitant immunosuppressant, N (%) Azathioprine 19 (16) Methotrexate 16 (14) Mycophenolate Mofetil 39 (33) Prednisolone dose at first RTX infusion, median (IQR) mg 10 (3–20) ESR at first RTX infusion, median (IQR) mm/hour 29 (15–57) BILAG index score at baseline, N (%) ≥1 A score 96 (82) No A score but ≥2 B scores 16 (14) BILAG domains at baseline, N (%) Grade A Grade B General 9 (8) 12 (10) Mucocutaneous 23 (20) 32 (27) Neurological 17 (15) 17 (15) Musculoskeletal 30 (26) 24 (20) Cardiorespiratory 6 (5) 13 (11) Gastrointestinal 6 (5) 0 (0) Ophthalmic 0 (0) 0 (0) Renal 34 (29) 0 (0) Haematology 11 (9) 12 (10) Global BILAG score, median (IQR) 21 (14–27) SLEDAI-2K score, median (IQR) 10 (6–14) SLICC Damage Index, median (IQR) 0 (0–1) ANA, antinuclear antibody; BILAG, British Isles Lupus Assessment Group; CYC, cyclophosphamide; dsDNA, double-stranded DNA; ESR, erythrocyte sedimentation rate; RNP, ribonucleic protein; RTX, rituximab; SLEDAI-2K, Systemic Lupus Erythematosus Disease Activity Index 2000; SLICC, Systemic Lupus International Collaborating Clinics (SLICC).

antibody; BILAG, British Isles Lupus Assessment Group; CYC, cyclophosphamide; dsDNA, double-stranded DNA; ESR, erythrocyte sedimentation rate; RNP, ribonucleic protein; RTX, rituximab; SLEDAI-2K, Systemic Lupus Erythematosus Disease Activity Index 2000; SLICC, Systemic Lupus International Collaborating Clinics (SLICC). Treatment characteristics Three hundred and eighteen cycles of rituximab were administered. Median (range) duration of response in rituximab responders for cycles 1–4 (C1–4) were 52 (26–423), 52 (26—299), 57 (27–184) and 50 (29–173) weeks, respectively. Concomitant cyclophosphamide was used in five patients who presented with life-threatening flare. Clinical and immunological response to first cycle rituximab In C1, there was a good overall clinical response to rituximab. Fifty-eight (50%) patients had major clinical response, 38 (32%) partial clinical response and 21 (18%) were non-responders. The median global BILAG scores had reduced from 21 (IQR 14–27) pre-rituximab to 8 (IQR 1–10) at 6 months; p<0.001. Responses in individual BILAG domains are shown in figure 1A. Although majority of domains improved, responses were more variable in the mucocutaneous and haematological domains. Mucocutaneous responses to rituximab have been described in detail previously.19 These long-term data showed a more consistent major response in lupus erythematosus non-specific lesions and oral ulcers, while non-response in chronic cutaneous lupus erythematosus (CCLE) (CCLE vs other lupus-specific lesions; p=0.022).

. Mucocutaneous responses to rituximab have been described in detail previously.19 These long-term data showed a more consistent major response in lupus erythematosus non-specific lesions and oral ulcers, while non-response in chronic cutaneous lupus erythematosus (CCLE) (CCLE vs other lupus-specific lesions; p=0.022). Figure 1 BILAG response and B-cell depletion following rituximab. (A) Majority of the individual domain improved post-rituximab although responses in the mucocutaneous and haematological domains were more varied. (B) Similar to the discovery cohort, a higher response rate was achieved in complete depletion compared with incomplete depletion groups; 93% versus 68%; p=0.011 in the validation cohort. (C) There was an incremental increase in the rates of B-cell depletion over three cycles of rituximab. ACLE, acute cutaneous lupus erythematosus; BILAG: British Isles Lupus Assessment Group; CCLE, chronic cutaneous lupus erythematosus; LENS, lupus erythematosus non-specific lesions. The median serum anti-dsDNA titre had reduced from 109 (IQR 16–300) IU/mL pre-rituximab to 32 (IQR 7–116) IU/mL at 6 months; p<0.001. Of 46 patients with low complement (C3 and/or C4) levels pre-rituximab, levels had normalised in 25/46 (54%) at 6 months. Predictors of major clinical response to first cycle rituximab Only B-cell depletion at 6 weeks increased the odds of BILAG response (major/partial) in multivariable analysis; adjusted imputed OR 13.93, 95% CI 3.11 to 62.37; p=0.001 (online supplementary table S2). 10.1136/annrheumdis-2017-211191.supp2Supplementary material 2

Predictors of major clinical response to first cycle rituximab Only B-cell depletion at 6 weeks increased the odds of BILAG response (major/partial) in multivariable analysis; adjusted imputed OR 13.93, 95% CI 3.11 to 62.37; p=0.001 (online supplementary table S2). 10.1136/annrheumdis-2017-211191.supp2Supplementary material 2 As there was a high degree of response to rituximab in this cohort, we analysed predictors for major clinical response separately in order to identify patients who would respond best to therapy. In imputed univariable analysis, only younger age was associated with major response to rituximab (OR 0.97, 95% CI 0.95 to 0.99; p=0.031). While in imputed multivariable model, younger age (OR 0.97, 95% CI 0.94 to 1.00; p=0.045) and B-cell depletion at 6 weeks post-rituximab (OR 3.22, 95% CI 1.24 to 8.33; p=0.016) increased the odds of major response to rituximab (table 2). Table 2 Multivariable analysis for predictors of major clinical response to first cycle rituximab

As there was a high degree of response to rituximab in this cohort, we analysed predictors for major clinical response separately in order to identify patients who would respond best to therapy. In imputed univariable analysis, only younger age was associated with major response to rituximab (OR 0.97, 95% CI 0.95 to 0.99; p=0.031). While in imputed multivariable model, younger age (OR 0.97, 95% CI 0.94 to 1.00; p=0.045) and B-cell depletion at 6 weeks post-rituximab (OR 3.22, 95% CI 1.24 to 8.33; p=0.016) increased the odds of major response to rituximab (table 2). Table 2 Multivariable analysis for predictors of major clinical response to first cycle rituximab No response/partial response n=59 Major clinical response n=58 Univariable OR (95% CI), p value (with multiple imputation) Multivariable OR (95% CI), p value (with multiple imputation) Age, mean (SD) years 43 (17) 37 (14) 0.97 (0.95 to 0.99), p=0.031 per year 0.97 (0.94 to 1.00), p=0.045 White, N (%) 43 (73) 37 (64) 1.53 (0.70 to 3.34), p=0.292 0.92 (0.34 to 2.47), p=0.870 Anti-dsDNA titres, mean (SD) IU/mL 147 (230) 142 (230) 1.00 (0.99 to 1.00), p=0.879 per unit 1.00 (0.99 to 1.00), p=0.632 Anti-ENA positivity, N (%) 40 (68) 38 (66) 0.91 (0.42 to 1.99), p=0.812 0.90 (0.37 to 2.22), p=0.821 Low C3 and/or C4 titres, N (%) 25 (42) 24 (41) 0.97 (0.46 to 2.04), p=0.937 1.14 (0.41 to 3.13), p=0.801 ESR, mean (SD) mm/hour* 40 (32) 41 (36) 1.00 (0.99 to 1.01), p=0.827 per unit – Concomitant S, N (%)† 41 (69) 35 (60) 0.67 (0.31 to 1.43), p=0.301 0.43 (0.17 to 1.09), p=0.075 Daily prednisolone dose, mean (SD) mg 13 (11) 16 (14) 1.02 (0.99 to 1.05), p=0.207 per mg 1.00 (0.97 to 1.04), p=0.713 Total BILAG score, mean (IQR) 21 (8) 24 (13) 1.03 (0.99 to 1.07), p=0.093 per point 1.02 (0.97 to 1.07), p=0.371 Total B-cell counts, mean (IQR)‡ 101 (95) 138 (150) 1.00 (1.00 to 1.01), p=0.161 per unit 1.00 (1.00 to 1.01), p=0.137 B-cell depletion at 6 weeks postrituximab, N (%) 29 (49) 39 (68) 2.10 (0.95 to 4.62), p=0.065 3.22 (1.24 to 8.33), p=0.016 *As high collinearity was observed between ESR and total B-cell counts, only the latter was included in the multivariable analysis.

.00 (1.00 to 1.01), p=0.161 per unit 1.00 (1.00 to 1.01), p=0.137 B-cell depletion at 6 weeks postrituximab, N (%) 29 (49) 39 (68) 2.10 (0.95 to 4.62), p=0.065 3.22 (1.24 to 8.33), p=0.016 *As high collinearity was observed between ESR and total B-cell counts, only the latter was included in the multivariable analysis. †Concomitant immunosuppressant was defined as either using methotrexate, azathioprine, mycophenolate mofetil and/or other disease modifying anti-rheumatic drugs but excluded anti-malarials. ‡count × 109 cells/L) for each subset multiplied by 1000 prior to analysis. BILAG, British Isles Lupus Assessment Group; C3/C4, complement 3 or 4; dsDNA, double-stranded DNA; ENA, extract nuclear antigen; ESR, erythrocyte sedimentation rate; IS, immunosuppressant. Validation of association between complete B-cell depletion and clinical response The published discovery cohort included 37 patients with SLE.2 In this validation cohort, 67 subsequent and consecutive patients (with B-cell data available) were analysed. Similar to the discovery cohort, higher response rate was achieved in complete depletion compared with incomplete depletion groups (93% vs 68%; p=0.011) in this validation cohort (figure 1B).

atients with SLE.2 In this validation cohort, 67 subsequent and consecutive patients (with B-cell data available) were analysed. Similar to the discovery cohort, higher response rate was achieved in complete depletion compared with incomplete depletion groups (93% vs 68%; p=0.011) in this validation cohort (figure 1B). While there was no difference at baseline, patients with complete B-cell depletion had significantly lower anti-dsDNA antibody titres at 14 weeks (p=0.030) and 26 weeks (p=0.041) versus those with incomplete depletion. In the former, C3 and C4 levels were not different at 14 weeks (p=0.064 and p=0.148, respectively) but were higher at 26 weeks (p=0.020 and p=0.022, respectively) compared with the latter group. There was no difference in anti-ENA antibodies between the two groups at 14 and 26 weeks; all p>0.10. Predictors for complete B-cell depletion to first cycle rituximab Data for B-cell subsets were available for 104 (89%) patients. In imputed univariable analysis, higher anti-dsDNA titre (OR 1.00, 95% CI 0.99 to 1.00; p=0.038), normal complement levels (OR 0.41, 95% CI 0.18 to 0.91; p=0.028) and lower pre-rituximab plasmablasts (OR 0.88, 95% CI 0.80 to 0.98; p=0.015) were associated with complete B-cell depletion. While in imputed multivariable model, only normal complement levels (OR 0.29, 95% CI 0.09 to 0.90; p=0.032) and lower pre-rituximab plasmablasts (OR 0.86, 95% CI 0.78 to 0.96; p=0.007) predicted complete B-cell depletion post-rituximab (online supplementary table S4).

15) were associated with complete B-cell depletion. While in imputed multivariable model, only normal complement levels (OR 0.29, 95% CI 0.09 to 0.90; p=0.032) and lower pre-rituximab plasmablasts (OR 0.86, 95% CI 0.78 to 0.96; p=0.007) predicted complete B-cell depletion post-rituximab (online supplementary table S4). B-cell depletion and associated serious infection As most of the serious infection episodes occurred in C1 and C2 (n=23 in 15 patients), we analysed the association between complete B-cell depletion and serious infection. After two cycles, there were no difference in the serious infection rates between complete and incomplete depletion groups (8/98 (8.2%) and 7/73 (9.6%), respectively; p=0.789). Plasmablast repopulation as a biomarker of relapse At 6 months, B-cells were detectable in 81% of the C1 responders. This time-point preceded all relapses. As the median of duration of response was 52 weeks, we divided the patients in this validation cohort (n=25 with B-cells data available) into two groups: (1) earlier relapse (≤12 months from first rituximab) and (2) later relapse (>12 months). A 12-month relapse time is clinically significant as it indicates that a 6-monthly retreatment may not be necessarily needed in these patients. Similar to the discovery cohort, the ROC indicated that a plasmablast count of >0.0008×109/L at 6 months yielded 73% (95% CI 45% to 92%) sensitivity and 90% (95% CI 56% to 99%) specificity in predicting earlier relapse; area under the curve of 0.86 (online supplementary figure S1). 10.1136/annrheumdis-2017-211191.supp1Supplementary Figure 1

Plasmablast repopulation as a biomarker of relapse At 6 months, B-cells were detectable in 81% of the C1 responders. This time-point preceded all relapses. As the median of duration of response was 52 weeks, we divided the patients in this validation cohort (n=25 with B-cells data available) into two groups: (1) earlier relapse (≤12 months from first rituximab) and (2) later relapse (>12 months). A 12-month relapse time is clinically significant as it indicates that a 6-monthly retreatment may not be necessarily needed in these patients. Similar to the discovery cohort, the ROC indicated that a plasmablast count of >0.0008×109/L at 6 months yielded 73% (95% CI 45% to 92%) sensitivity and 90% (95% CI 56% to 99%) specificity in predicting earlier relapse; area under the curve of 0.86 (online supplementary figure S1). 10.1136/annrheumdis-2017-211191.supp1Supplementary Figure 1 Of the patients with plasmablasts >0.0008×109/L at 6 months, relapse rates within the next 26 and 52 weeks were 90% and 100%, respectively. While of the patients with plasmablasts ≤0.0008×109/L at 6 months, relapse rates within the next 26 and 52 weeks were 33% and 73%, respectively (figure 2A).

10.1136/annrheumdis-2017-211191.supp1Supplementary Figure 1 Of the patients with plasmablasts >0.0008×109/L at 6 months, relapse rates within the next 26 and 52 weeks were 90% and 100%, respectively. While of the patients with plasmablasts ≤0.0008×109/L at 6 months, relapse rates within the next 26 and 52 weeks were 33% and 73%, respectively (figure 2A). Figure 2 2NDNR to rituximab and efficacy of alternative humanised anti-CD20 antibodies. (A) In this validation cohort, detection of plasmablasts >0.0008×109/L at 6 months predicted earlier relapse. (B) The phenomenon 2NDNR was associated with anti-rituximab antibody. The dotted red line represents normal cut-off of the test. (C) The Global BILAG score and CD20+ B-cells are plotted for each patient. The black line in the CD20+ B-cells figure represents the median. (D) An example of a case where proteinuria was normalised following a switch to ocrelizumab. ‘RR’ represents 2x infusions of rituximab, ‘R’ represents a single infusion as the patient cannot not complete the second due to severe infusion reaction and ‘OO’ represents 2x infusions of ocrelizumab. The total B-cell counts were transformed to natural log. 2NDNR, secondary non-depletion non-response; BILAG, British Isles Lupus Assessment Group. There were no differences in anti-dsDNA titres, total BILAG score and memory B-cells at 6 months between the earlier versus later relapse groups, p=0.475, p=0.985 and p=0.414, respectively.

Figure 2 2NDNR to rituximab and efficacy of alternative humanised anti-CD20 antibodies. (A) In this validation cohort, detection of plasmablasts >0.0008×109/L at 6 months predicted earlier relapse. (B) The phenomenon 2NDNR was associated with anti-rituximab antibody. The dotted red line represents normal cut-off of the test. (C) The Global BILAG score and CD20+ B-cells are plotted for each patient. The black line in the CD20+ B-cells figure represents the median. (D) An example of a case where proteinuria was normalised following a switch to ocrelizumab. ‘RR’ represents 2x infusions of rituximab, ‘R’ represents a single infusion as the patient cannot not complete the second due to severe infusion reaction and ‘OO’ represents 2x infusions of ocrelizumab. The total B-cell counts were transformed to natural log. 2NDNR, secondary non-depletion non-response; BILAG, British Isles Lupus Assessment Group. There were no differences in anti-dsDNA titres, total BILAG score and memory B-cells at 6 months between the earlier versus later relapse groups, p=0.475, p=0.985 and p=0.414, respectively. Retreatment of first cycle non-responders In RA, we showed that retreatment of initial non-responders with incomplete B-cell depletion led to improved response rate in C2.20 Of the 21 patients who were C1 non-responders, nine were retreated with rituximab. The domains that persisted at grade A/B in C1 were mucocutaneous (n=4), musculoskeletal (n=3), renal (n=2) and haematology (n=3). After retreatment, none of these patients responded. Additionally, four patients had clinical features that were suggestive of immunogenicity.

responders, nine were retreated with rituximab. The domains that persisted at grade A/B in C1 were mucocutaneous (n=4), musculoskeletal (n=3), renal (n=2) and haematology (n=3). After retreatment, none of these patients responded. Additionally, four patients had clinical features that were suggestive of immunogenicity. Retreatment of first cycle responders Of the 96 patients who were C1 responders, 77 (with complete data on 72) were retreated on clinical relapse. Of these, 61/72 (85%) responded in C2 (figure 3). Numerically higher rate of B-cell depletion was achieved in C2 compared with C1 (68% versus 58%, respectively; p=0.206) and depletion improved over subsequent cycle, C3 versus C1 (79% vs 58% respectively; p=0.022) (figure 1C). Figure 3 Efficacy of repeat cycles with rituximab in systemic lupus erythematosus. There was a high rate of initial clinical response to rituximab in this cohort, 96/117 (82%). Seventy-seven responders who had clinical relapse were retreated in C2. Of these, 61/72 (85%) continued to response in C2. Of the C2 non-responders, 9/11 met 2NDNR criteria. Five were switched to ocrelizumab/ofatumumab resulted in depletion and response in all. 2NDNR, secondary non-depletion and non-response; C1, cycle 1. Twelve out of 38 patients who were C1 partial responders were retreated at 6 months. Of these, major clinical response was achieved in 10/12 (83%) in C2. One patient had worsening of arthritis, while another had 2NDNR in C2.

Figure 3 Efficacy of repeat cycles with rituximab in systemic lupus erythematosus. There was a high rate of initial clinical response to rituximab in this cohort, 96/117 (82%). Seventy-seven responders who had clinical relapse were retreated in C2. Of these, 61/72 (85%) continued to response in C2. Of the C2 non-responders, 9/11 met 2NDNR criteria. Five were switched to ocrelizumab/ofatumumab resulted in depletion and response in all. 2NDNR, secondary non-depletion and non-response; C1, cycle 1. Twelve out of 38 patients who were C1 partial responders were retreated at 6 months. Of these, major clinical response was achieved in 10/12 (83%) in C2. One patient had worsening of arthritis, while another had 2NDNR in C2. Of the 11 patients who were C2 non-responders, nine met 2NDNR criteria. Therefore, the incidence of 2NDNR in this cohort was 9/77 (12%). In C3, another two patients had 2NDNR. Association of 2NDNR with antirituximab antibody Post-rituximab sera for 5/9 patients with 2NDNR were tested for anti-rituximab antibodies. Of these, all 5/5 (100%) were tested positive. In contrast, of the 16 patients who were C2 responders, 9/16 (56%) were also tested positive for anti-rituximab antibodies. The median anti-rituximab levels were higher in the former, 562 (IQR 394–9670) AU/mL compared with the latter, 217 (IQR 0–409) AU/mL; p=0.024 (figure 2B).

ese, all 5/5 (100%) were tested positive. In contrast, of the 16 patients who were C2 responders, 9/16 (56%) were also tested positive for anti-rituximab antibodies. The median anti-rituximab levels were higher in the former, 562 (IQR 394–9670) AU/mL compared with the latter, 217 (IQR 0–409) AU/mL; p=0.024 (figure 2B). Factors associated with 2NDNR Risk factors for 2NDNR were lack of concomitant immunosuppressant (p=0.023) and higher pre-rituximab plasmablasts (p<0.001) (table 3). Concomitant corticosteroid dose, duration of response in C1, clinical response category in C1, pre-rituximab global BILAG score, pre-rituximab naïve and memory B-cells were not associated with 2NDNR; all p>0.10. Table 3 Factors associated with secondary non-depletion non-response to rituximab (2NDNR) Characteristics prior to rituximab retreatment Continued to respond (n=61) 2NDNR (n=9) p Value Concomitant IS, N (%) 41 (67) 2 (22) 0.023 Prednisolone, median (IQR) mg 5 (0–10) 5 (0–17.5) 0.729 Duration of response, median (IQR) weeks 50 (36–107) 62 (52–164) 0.239 Total BILAG score, median (IQR) 16 (12–21) 24 (12–27) 0.209 Partial clinical response in cycle 1, N (%) 24 (39) 3 (33) 0.731 Naïve B-cells, median (IQR) 109 cells/L 0.0349 (0.0071–0.0735) 0.0620 (0.0101–0.0950) 0.296 Memory B-cells, median (IQR) x 109/L 0.0019 (0.0010–0.0047) 0.0090 (0.0054–0.0394) 0.175 Plasmablasts, median (IQR) x 109/L 0.0011 (0.0004–0.0036) 0.0086 (0.0052–0.0227) <0.001 *NDNR, secondary non-depletion and non-response; IS, immunosuppressant.

n (IQR) 109 cells/L 0.0349 (0.0071–0.0735) 0.0620 (0.0101–0.0950) 0.296 Memory B-cells, median (IQR) x 109/L 0.0019 (0.0010–0.0047) 0.0090 (0.0054–0.0394) 0.175 Plasmablasts, median (IQR) x 109/L 0.0011 (0.0004–0.0036) 0.0086 (0.0052–0.0227) <0.001 *NDNR, secondary non-depletion and non-response; IS, immunosuppressant. Efficacy of switching to humanised anti-CD20 antibodies Following 2NDNR, treatment for five patients were switched to humanised anti-CD20 mAbs (3=ocrelizumab and 2=ofatumumab). Post-treatment, complete depletion of CD20+ cells were achieved in 4/5 patients, while the remaining one had substantially low counts (0.0016×109/L). The median global BILAG scores had reduced from 24 (IQR 18–45) pre-treatment to 1 (IQR 0–8) post-treatment; p=0.008 (figure 2C). The individual BILAG response is shown in figure 2D and described in online supplementary table S5. One patient with class IV-G (active with moderate scarring) who had progressed into end-stage renal failure was treated with ofatumumab, mainly for severe thrombocytopaenia with a view for renal transplantation preparation. Post-treatment, her platelet had normalised from 45×109/L (pre-treatment), renal parameters were stable and she successfully underwent live donor renal transplantation.

ressed into end-stage renal failure was treated with ofatumumab, mainly for severe thrombocytopaenia with a view for renal transplantation preparation. Post-treatment, her platelet had normalised from 45×109/L (pre-treatment), renal parameters were stable and she successfully underwent live donor renal transplantation. Discussion The clinical challenges for the use of rituximab in SLE include defining subgroups of patients likely to respond to the initial and subsequent cycles and optimal repeat treatment strategy. By capturing data of all patients with SLE who were treated with rituximab in this largest reported cohort, as well as long-term follow-up, this study offers insights into pragmatic use of rituximab and has implications for the future development of targeted therapies.

uent cycles and optimal repeat treatment strategy. By capturing data of all patients with SLE who were treated with rituximab in this largest reported cohort, as well as long-term follow-up, this study offers insights into pragmatic use of rituximab and has implications for the future development of targeted therapies. In this study, the only consistent predictor of any (and major) clinical response to rituximab is B-cell depletion (as measured using HSFC) at 6 weeks post-rituximab, which we have now validated in an independent cohort. This underlines the immunomodulatory action of rituximab in correcting autoimmune B-cell function and normalising autoantibody titres and complement levels without increasing the risk of severe infection. From treatment stratification perspective, our data support the rationale for B-cell monitoring during therapy. Thus, prior to rituximab, by assessing patients for low complement levels and higher plasmablasts, treatment modification can be employed to improve depletion, either by increasing the dose or adding an extra infusion, as we previously showed in RA.21 At 6 weeks post-rituximab, complete depletion is a marker of good response to therapy. For those with incomplete depletion, close monitoring is required. At 6 months post-rituximab, repopulation of plasmablasts of >0.0008×109/L increases the risk of clinical relapse within the following 6 months. Therefore, these patients can be considered for early retreatment in order to reduce the higher burden of B-cell numbers and enhance depletion in the subsequent cycle. Importantly, for those with plasmablasts of ≤0.0008×109/L at 6 months, monitoring for clinical relapse would appear an acceptable strategy.

ng 6 months. Therefore, these patients can be considered for early retreatment in order to reduce the higher burden of B-cell numbers and enhance depletion in the subsequent cycle. Importantly, for those with plasmablasts of ≤0.0008×109/L at 6 months, monitoring for clinical relapse would appear an acceptable strategy. Regardless of response, about 12% subsequently developed 2NDNR in C2. This phenomenon is associated with rituximab anti-drug antibodies. However, measuring anti-rituximab antibody alone is not enough to identify patients as 2NDNR as over half of the patients who were tested positive responded in that particular cycle. Instead, clinical features, that is, severe infusion reaction and non-response and measuring B-cells, are more meaningful. Lack of concomitant oral immunosuppressant and higher pre-rituximab plasmablasts predicted 2NDNR. Oral immunosuppressant use was decided at physician discretion, but our data suggest they might have a role in preventing immunogenicity. The exact mechanism for the association with plasmablast number is unknown, but plasmablasts are markers for overall B-cell activation. Following initial depletion with rituximab, B-cell-activating factor levels increase and promote the formation of plasmablasts.22 This early increase in plasmablasts enhances the formation of follicular T-helper cells, thus creating a positive feedback loop that perpetuates antibody-driven inflammation and may explain why some patients become refractory to rituximab in SLE.23

ing factor levels increase and promote the formation of plasmablasts.22 This early increase in plasmablasts enhances the formation of follicular T-helper cells, thus creating a positive feedback loop that perpetuates antibody-driven inflammation and may explain why some patients become refractory to rituximab in SLE.23 Following 2NDNR to rituximab, switching to humanised anti-CD20 mAbs restores depletion and response in SLE. Ocrelizumab and ofatumumab are both type 1 anti-CD20 mAbs. The primary endpoint was met in ocrelizumab-treated groups in RA trials24 and was investigated in SLE.25 However, development in these indications was halted after an increase in opportunistic infections, some of which fatal were reported.26 All three patients in our study had major clinical responses and prolonged remission for over 5-year period post-ocrelizumab. Ofatumumab is licenced for resistant chronic lymphocytic leukaemia and has demonstrated efficacy in RA.27 Both patients in our study responded well to ofatumumab included one who achieved complete depletion for the first time from B-cell depleting therapy. Additionally, a few case series have recently reported on its efficacy in extrarenal and refractory lupus nephritis.28 29 Alternatively, other anti-CD20 agents with enhanced antibody-dependent cellular cytotoxicity may be more effective in SLE. In vitro obinutuzumab demonstrated enhanced depletion was achieved with this type 2 mAb, compared with rituximab.30

ently reported on its efficacy in extrarenal and refractory lupus nephritis.28 29 Alternatively, other anti-CD20 agents with enhanced antibody-dependent cellular cytotoxicity may be more effective in SLE. In vitro obinutuzumab demonstrated enhanced depletion was achieved with this type 2 mAb, compared with rituximab.30 This study has several limitations. First, an interobserver variability could have occurred in BILAG assessments due to the lengthy follow-up duration and a cohort that was highly heterogeneous in lupus manifestations. However, the BILAG scores reflected the clinician’s intention-to-treat, and the patients were managed in a dedicated single centre, thus allowing for consistency in assessment. Second, B-cells and laboratory data were missing in some cases. As these were deemed missing at random, multiple imputation was used to reduce potential bias in parameter estimation as well as enhancing generalisability of the results. Next, concomitant therapy with immunosuppressant were used in more than 60% of the patients, thus efficacy could not be attributed to rituximab alone. Lastly, the lack of control group limits interpretation of efficacy and safety of rituximab.

n parameter estimation as well as enhancing generalisability of the results. Next, concomitant therapy with immunosuppressant were used in more than 60% of the patients, thus efficacy could not be attributed to rituximab alone. Lastly, the lack of control group limits interpretation of efficacy and safety of rituximab. In conclusion, treatment with anti-CD20 agents can be guided by B-cell monitoring with the aim of achieving complete depletion. About one in eight patients with SLE lose depletion on repeat cycles of rituximab regardless of prior response and secondary non-depletion is associated with anti-rituximab antibodies. Concomitant oral immunosuppressant may help to prevent this. If 2NDNR occurs, switching to humanised anti-CD20 mAbs restores depletion and response. Therefore, alternative anti-CD20 antibodies may be more consistently effective in SLE treatment and several ongoing trials are addressing these issues. The authors would like to thank Elizabeth M Hensor for statistical advice, clinicians, pharmacist, study coordinator and lab technicians at the Leeds Connective Tissue Disease and Vasculitis Clinic particularly Mike Martin, Jacqueline Andrews, Maya Buch, Colin Pease, Shouvik Dass, Sinisa Savic, John Bamford, Paul Beirne, Mark Goodfield, Tina Hawkins, Huma Cassamoali, Diane Corscadden and Katie Mbara for their substantial contribution in the acquisition of the data.

Connective Tissue Disease and Vasculitis Clinic particularly Mike Martin, Jacqueline Andrews, Maya Buch, Colin Pease, Shouvik Dass, Sinisa Savic, John Bamford, Paul Beirne, Mark Goodfield, Tina Hawkins, Huma Cassamoali, Diane Corscadden and Katie Mbara for their substantial contribution in the acquisition of the data. Contributors: MYMY, PE and EMV: substantial contributions to the conception or design of the work, or the acquisition, analysis or interpretation of data, drafting the work or revising it critically for important intellectual content, final approval of the version published and agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. DS, YME-S, ED and ACR: substantial contributions to the conception or design of the work, or the acquisition, analysis or interpretation of data, drafting the work or revising it critically for important intellectual content and final approval of the version published. Funding: This research was funded/supported by the National Institute for Health Research (NIHR) and NIHR Leeds Biomedical Research Centre based at Leeds Teaching Hospitals NHS Trust; and NIHR Research Grants (DRF-2014-07-155) and (CS-2013-13-032). The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR or the Department of Health.

National Institute for Health Research (NIHR) and NIHR Leeds Biomedical Research Centre based at Leeds Teaching Hospitals NHS Trust; and NIHR Research Grants (DRF-2014-07-155) and (CS-2013-13-032). The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR or the Department of Health. Competing interests: EMV is an NIHR Clinician Scientist. He has received honoraria and research grant support from Roche, GSK and AstraZeneca. PE has received consultant fees from BMS, Abbott, Pfizer, MSD, Novartis, Roche and UCB. He has received research grants paid to his employer from AstraZeneca, Abbott, BMS, Pfizer, MSD and Roche. Patient consent: The study does not contain any personal medical information about an identifiable living individual, thus patient consent is not required. Ethics approval: The use of rituximab, ofatumumab and ocrelizumab were all approved by Leeds Teaching Hospitals NHS Trust Drug and Therapeutic Committee. Analysis of samples for antirituximabantibody was approved by the Leeds (East) Research Ethics Committee (REC), 10/H1306/88, and the committee confirmed that other aspects of the study did not require ethical approval in accordance with the UK National Health Service REC guidelines. Provenance and peer review: Not commissioned; externally peer reviewed. Correction notice: This article has been corrected since it published Online First. The abstract has been corrected.

Introduction Primary Sjögren’s syndrome (PSS) is characterised by focal lymphocytic infiltration of exocrine glands leading to profound dryness. It is often accompanied by systemic manifestations and high levels of fatigue. B cells are considered to have a central role in pathogenesis,1 and two small randomised controlled trials (RCTs) of the anti-CD20 B-cell-depleting agent rituximab suggested benefits in PSS.2 3 Rituximab may also have effects on interleukin-17-producing mast cells and on a CD20-positive T cell subset.4 5 Despite this, French (TEARS) and British (TRACTISS) phase III RCTs failed to demonstrate an effect on primary endpoints based on patient-reported visual analogue scales (VAS).6 7 Potential explanations for these disappointing findings include the lack of patient stratification, insufficient tissue depletion of B cells and the choice and timing of primary outcome.

CTISS) phase III RCTs failed to demonstrate an effect on primary endpoints based on patient-reported visual analogue scales (VAS).6 7 Potential explanations for these disappointing findings include the lack of patient stratification, insufficient tissue depletion of B cells and the choice and timing of primary outcome. The requirement for new and validated outcome measures for PSS led to the development of the European Sjögren’s Syndrome Patient Reported Index (ESSPRI) and a physician-assessed systemic disease activity index (European League Against Rheumatism Sjögren’s Syndrome Disease Activity Index (ESSDAI)).8 These are a welcome advance, but certain limitations suggest that additional objective outcome measures/biomarkers would be desirable. Use of the ESSDAI, for example, requires a minimum threshold for trial entry that excludes a large proportion of patients. Other outcome measures include salivary flow rates, although these are subject to issues of standardisation and diurnal variation,9 and histological examination of salivary gland biopsies, which may provide mechanistic information but is invasive.10 11 Salivary gland ultrasound (SGUS) is readily available, non-invasive and shows reasonable sensitivity and good specificity for the diagnosis of PSS.12–14 In PSS, glandular echogenicity is altered and there is loss of homogeneity due to the presence of multiple hypoechoic or anechoic areas, as well as hyperechoic bands. Loss of definition of the glandular border may also be observed. A single-site substudy of SGUS in TEARS showed that a greater number of patients had improvement in parotid gland echostructure at 24 weeks after rituximab compared with placebo.15 Echostructure was assessed on a 0–4 scale that graded the presence of hypoechoic areas as well as hyperechoic bands. SGUS is, however, an operator-dependent technique, and its utility in a multicentre study is uncertain. Here we report the results of a multiobserver, multicentre SGUS substudy of TRACTISS over a longer therapeutic timeframe.

ssed on a 0–4 scale that graded the presence of hypoechoic areas as well as hyperechoic bands. SGUS is, however, an operator-dependent technique, and its utility in a multicentre study is uncertain. Here we report the results of a multiobserver, multicentre SGUS substudy of TRACTISS over a longer therapeutic timeframe. Methods The TRACTISS study has been previously described.6 Briefly, 133 patients with PSS were randomised 1:1 to 1000 mg rituximab or placebo given at weeks 0, 2, 24 and 26. Patients and clinicians were blind to the randomised allocation. The primary outcome (30% reduction in either oral dryness or fatigue VAS) was assessed at week 48. Methylprednisolone 100 mg was given prior to each infusion of rituximab or placebo. Subjects could consent to an optional SGUS substudy, with assessments at baseline and weeks 16 and 48. The prespecified substudy primary outcome was total ultrasound score (TUS, range 0–11; table 1). Normal salivary gland echogenicity was defined through similarity with the thyroid. The consistency domain scored the extent of heterogeneity introduced by the presence of hypoechoic areas. The definition domain addressed whether the posterior glandular border was normally visible or else incompletely defined or not possible to define. The hypoechoic foci size domain categorised the size of the glandular hypoechoic lesions that were most typical for that patient. Imaging followed a standard sequence including both transverse and longitudinal views of both parotid and submandibular glands, with data recorded by the sonographer on a study proforma. Additional information was collected for each of the four major salivary glands on vascularity of the gland parenchyma assessed by power Doppler, gland echogenicity (normal, heterogenous or hypoechoic), gland margins (well or ill-defined), approximate hypoechoic foci number (0, 1–5, 5–9 and >10), hypoechoic foci size (<3, 3–7 and >8 mm), as well as domains capturing lymph node abnormalities.

lands on vascularity of the gland parenchyma assessed by power Doppler, gland echogenicity (normal, heterogenous or hypoechoic), gland margins (well or ill-defined), approximate hypoechoic foci number (0, 1–5, 5–9 and >10), hypoechoic foci size (<3, 3–7 and >8 mm), as well as domains capturing lymph node abnormalities. Table 1 Domains of the total ultrasound score Domain Description Score Echogenicity Normal 0 Hypoechoic 1 Consistency Normal 0 Mild heterogeneity 1 Evident honeycombed 2 Gross multifocal 3 Definition Normal 0 Moderately defined 1 Ill-defined 2 Glands involved None 0 Parotids or submandibular glands 1 All glands 2 Hypoechoic foci size None 0 Small 2–5 mm 1 Large 5–8 mm non-vascular 2 Over 8 mm ± vascular 3 Total 0–11 ESSPRI score was calculated as the mean of 0–10 scales for dryness, fatigue and limb pain. The ESSDAI score was scored by the local investigator. Unstimulated whole salivary flow was collected over 15 min, and stimulated whole salivary flow over 10 min following application of citric acid with a cotton swab to the lateral borders of the tongue every 60 s. TUS was modelled using mixed effects linear regression, including baseline score, patient age, disease duration and time point. Odds of domain improvement were modelled by repeated-measures logistic regression, including baseline score, age, disease duration and time point. Descriptive summary statistics, scatterplots and boxplots were produced to explore and summarise the data.

ding baseline score, patient age, disease duration and time point. Odds of domain improvement were modelled by repeated-measures logistic regression, including baseline score, age, disease duration and time point. Descriptive summary statistics, scatterplots and boxplots were produced to explore and summarise the data. Results In total, 66 patients (49.6%) from the total study population consented to SGUS, and 52 (39.1%; n=26 rituximab and n=26 placebo) patients from nine centres completed the baseline and at least one follow-up visit. Of these 52 patients, 43 (83%) completed all three visits. There were no apparent differences in relevant characteristics between those consenting and not consenting to the substudy (online supplementary table S1). The two arms of the substudy were also similar (table 2), although TUS in the rituximab arm was on average one point greater. 10.1136/annrheumdis-2017-212268.supp1Supplementary file 1 Figure 1 illustrates the baseline-adjusted values of TUS over time, modelling the change from baseline at each time point. Estimated baseline-adjusted TUS at week 16 was 6.2 (95% CI 5.4 to 7.0) for placebo and 5.0 (95% CI 4.4 to 5.6) for rituximab, and at week 48, 6.1 (95% CI 5.5 to 6.6) and 4.8 (95% CI 4.2 to 5.4), respectively. Estimated between-group differences (rituximab-placebo) in baseline-adjusted TUS were −1.2 (95% CI −2.1 to −0.3; P=0.0099) and −1.2 (95% CI −2.0 to –0.5; P=0.0023) at weeks 16 and 48, respectively. Table 2 Selected baseline characteristics of subjects with both baseline and follow-up data in salivary gland ultrasound substudy

Figure 1 illustrates the baseline-adjusted values of TUS over time, modelling the change from baseline at each time point. Estimated baseline-adjusted TUS at week 16 was 6.2 (95% CI 5.4 to 7.0) for placebo and 5.0 (95% CI 4.4 to 5.6) for rituximab, and at week 48, 6.1 (95% CI 5.5 to 6.6) and 4.8 (95% CI 4.2 to 5.4), respectively. Estimated between-group differences (rituximab-placebo) in baseline-adjusted TUS were −1.2 (95% CI −2.1 to −0.3; P=0.0099) and −1.2 (95% CI −2.0 to –0.5; P=0.0023) at weeks 16 and 48, respectively. Table 2 Selected baseline characteristics of subjects with both baseline and follow-up data in salivary gland ultrasound substudy Placebo (n=26) Rituximab (n=26) All (n=52) Age (years) 57.4 (11.1) 56.7 (10.92) 57.1 (10.91) Years since diagnosis 6.6 (5.67) 5.38 (4.82) 6.0 (5.25) ≥10 years since diagnosis, n (%) 6 (23.1) 4 (15.4) 10 (19.2) Female sex, n (%) 23 (88.5) 25 (96.2) 48 (92.3) Current medications (prior to randomisation) Pilocarpine, n (%) 1 (3.8) 4 (15.4) 5 (9.6) Hydroxychloroquine, n (%) 13 (50.0) 15 (57.7) 28 (53.8) Corticosteroids, n (%) 6 (23.1) 2 (7.7) 8 (15.4) NSAIDS: n (%) 7 (26.9) 5 (19.2) 12 (23.1) Unstimulated salivary flow (mL/15 min) 1.4 (2.34) 0.8 (0.71) 1.1 (1.72) Stimulated salivary flow (mL/10 min) 3.8 (4.08) 3.7 (5.51) 3.7 (4.82) IgG (g/L) 17.2 (7.67) 17.8 (6.02) 17.5 (6.82) IgA (g/L) 3.7 (2.87) 3.0 (1.0) 3.3 (2.14) IgM (g/L) 1.2 (0.64) 1.4 (0.65) 1.28 (0.64) Anti-Ro autoantibody positive, n (%) 26 (100) 25 (96.2) 51 (98.1) Reduced C4, n (%) 4 (15.4) 4 (15.4) 8 (15.4) Visual analogue scales (average over last two weeks, mm; 100=severe, except global) Fatigue 74.5 (13.46) 67.0 (18.22) 70.8 (16.30) Oral dryness 75.6 (15.13) 73.8 (13.30) 74.7 (14.14) Ocular dryness 64.7 (23.25) 65.7 (19.25) 65.2 (21.09) Overall dryness 73.4 (15.64) 71.3 (13.17) 72.4 (14.36) Joint pain 56.4 (28.40) 47.2 (27.21) 51.8 (27.93) Global assessment (100=PSS very active) 73.4 (14.08) 62.2 (18.90) 67.8 (17.45) ESSPRI (10=maximal symptom severity) 6.7 (1.63) 6.4 (1.64) 6.6 (1.64) ESSDAI (123=maximal disease activity) 6.8 (3.82) 5.1 (4.55) 6.0 (4.24) ESSDAI glandular domain, n (%) No activity 17 (65.4) 22 (84.6) 39 (75.0) Low activity 8 (30.8) 3 (11.5) 11 (21.2) Moderate activity 1 (3.8) 1 (3.8) 2 (3.8) TUS 5.02 (3.06) 6.5 (2.04) 5.9 (2.65) TUS domains Echogenicity 0.5 (0.51) 0.8 (0.43) 0.7 (0.48) Consistency 1.3 (1.00) 1.5 (0.91) 1.4 (0.95) Definition 0.8 (0.83) 1.3 (0.74) 1.0 (0.82) Glands involved 1.5 (0.81) 1.9 (0.43) 1.7 (0.67) Hypoechoic foci size 1.0 (0.68) 1.1 (0.48) 1.1 (0.58) Values are mean and SD unless otherwise stated.

6.5 (2.04) 5.9 (2.65) TUS domains Echogenicity 0.5 (0.51) 0.8 (0.43) 0.7 (0.48) Consistency 1.3 (1.00) 1.5 (0.91) 1.4 (0.95) Definition 0.8 (0.83) 1.3 (0.74) 1.0 (0.82) Glands involved 1.5 (0.81) 1.9 (0.43) 1.7 (0.67) Hypoechoic foci size 1.0 (0.68) 1.1 (0.48) 1.1 (0.58) Values are mean and SD unless otherwise stated. ESSDAI, European League Against Rheumatism Sjögren’s Syndrome Disease Activity Index; ESSPRI, European Sjögren’s Syndrome Patient Reported Index; NSAID, non-steroidal anti-inflammatory drugs; PSS, primary Sjögren’s syndrome; TUS, total ultrasound score. Figure 1 Baseline-adjusted total ultrasound score (TUS) at follow-up. Mean baseline-adjusted TUS, and between-group differences at weeks 16 and 48. Data modelled using a covariance pattern mixed model, with the baseline value fitted as a fixed effect. Values presented are least-squares means and 95% CIs for the two groups, and the differences between the groups. PLC, placebo; RTX, rituximab.

baseline-adjusted TUS, and between-group differences at weeks 16 and 48. Data modelled using a covariance pattern mixed model, with the baseline value fitted as a fixed effect. Values presented are least-squares means and 95% CIs for the two groups, and the differences between the groups. PLC, placebo; RTX, rituximab. For each TUS domain, we fitted a repeated-measures logistic regression to model the odds of a response in the rituximab arm (defined as ≥1-point improvement) as a function of the baseline score, age category, disease duration and time point. Glandular definition was the only domain to show statistically significant improvement with an OR of 6.8 (95% CI 1.1 to 43.0; P=0.043) at week 16 and 10.3 (95% CI 1.0 to 105.9; P=0.050) at week 48 (online supplementary table S2). No difference between rituximab and placebo was observed in any of the additional exploratory ultrasound parameters collected, with the exception of gland margin scores which showed deterioration in the placebo group (mean sum of scores over all glands increasing from 1.8 (SD 1.95) at baseline to 2.4 (SD 1.89) at 48 weeks compared with 2.3 (SD 1.83) to 2.4 (SD 1.97) in the rituximab group).

ditional exploratory ultrasound parameters collected, with the exception of gland margin scores which showed deterioration in the placebo group (mean sum of scores over all glands increasing from 1.8 (SD 1.95) at baseline to 2.4 (SD 1.89) at 48 weeks compared with 2.3 (SD 1.83) to 2.4 (SD 1.97) in the rituximab group). Improvement of ≥1 point in TUS, compared with no improvement or worsening, was not associated with improvement in unstimulated or stimulated salivary flow rates, ESSPRI score or dryness domain VAS at weeks 16 or 48, in the whole population or when analysing the rituximab arm alone. No associations were observed with ≥1-point improvement in either the glandular definition or hypoechoic foci size domains. TUS did not correlate with total ESSDAI score, the ESSDAI glandular domain or salivary flow rates at any time point, either in the whole population or the rituximab arm. Baseline TUS was not correlated with improvement in salivary flow rates, ESSPRI or oral dryness VAS at either week 16 or 48 in the rituximab arm (data not shown).

correlate with total ESSDAI score, the ESSDAI glandular domain or salivary flow rates at any time point, either in the whole population or the rituximab arm. Baseline TUS was not correlated with improvement in salivary flow rates, ESSPRI or oral dryness VAS at either week 16 or 48 in the rituximab arm (data not shown). Discussion We demonstrated a statistically significant improvement in TUS after rituximab compared with placebo. While this observation is similar to that in the TEARS substudy, there are a number of key differences. First, in TRACTISS rituximab was given at baseline and then again at 6 months, with a longer follow-up to 48 weeks. Second, the TRACTISS substudy was larger, multicentre and multiobserver. The ability of ultrasound to detect changes in this setting is important in encouraging further development of this tool. Third, TRACTISS used a composite SGUS score. Fourth and related to the last point, the number and size of hypoechogenic foci showed no change in TRACTISS, in contrast to the TEARS study.

tiobserver. The ability of ultrasound to detect changes in this setting is important in encouraging further development of this tool. Third, TRACTISS used a composite SGUS score. Fourth and related to the last point, the number and size of hypoechogenic foci showed no change in TRACTISS, in contrast to the TEARS study. The pathological correlate of the hypoechoic areas observed on ultrasound in PSS is uncertain. In TEARS, there was a correlation between histological focus score and SGUS score, suggesting that hypoechoic areas represent areas of inflammatory cell infiltrate.16 Furthermore, both high baseline SGUS score and high numbers of infiltrating B cells were predictive of non-response.17 18 However, opposite findings on B cell infiltration and rituximab responsiveness have been reported by Delli et al,19 and in a cohort of patients with suspected PSS there was only a modest agreement between the same SGUS score and biopsy.13 Therefore, it remains possible that the highest grades of hypoechoic lesions might reflect damage as well as inflammation in a subset of patients, explaining why we observed no change in their size or number. Our results suggest that glandular definition was an important domain driving change in TUS. While there is a pragmatic attractiveness in simplified scores focusing on hypoechoic areas for diagnosis,20 our data encourage the collection of a wider range of features/domains in clinical trials as there is yet much to learn about the responsiveness of US to effective treatments in PSS.

omain driving change in TUS. While there is a pragmatic attractiveness in simplified scores focusing on hypoechoic areas for diagnosis,20 our data encourage the collection of a wider range of features/domains in clinical trials as there is yet much to learn about the responsiveness of US to effective treatments in PSS. The clinical significance of our findings is uncertain. TRACTISS did not meet its primary endpoint,6 and no association between TUS improvement and salivary flow was found. We also found no apparent inverse association between salivary flow rates and TUS at baseline, in contrast to previous cross-sectional studies, which may reflect our small sample size given that previously reported correlations were only fair to moderate.21 22 Furthermore, the improvement in the glandular definition domain was only of marginal statistical significance. We used a novel composite score, designed to be comprehensive but also pragmatic, but which predated the EULAR pSS working group reference atlas.23 Other limitations include the small number of subjects and the multiplicity of statistical comparisons, for which we did not adjust our nominal significance levels. Although the sonographers in this study were experienced in SGUS, ultrasound machines were not standardised between centres, and some domains, especially the definition domain, can be difficult to assess. Intraobserver and interobserver reliability was not studied and could have impacted our findings; further standardisation of SGUS in PSS is urgently required. Arguably, however, the ability to distinguish treatment arms despite such standardisation may increase the relevance of our findings.

can be difficult to assess. Intraobserver and interobserver reliability was not studied and could have impacted our findings; further standardisation of SGUS in PSS is urgently required. Arguably, however, the ability to distinguish treatment arms despite such standardisation may increase the relevance of our findings. There is good reason to believe that rituximab monotherapy may stimulate new autoimmune B cells through elevation in BLyS levels24 and may be inefficient at depleting tissue B cells.25 The fact that we observed a difference in TUS between study arms despite these limitations encourages further research on B cell depletion therapy in PSS, including use of combination therapies,26 and on SGUS as an imaging biomarker. Handling editor: Tore K Kvien Contributors: BAF drafted the manuscript and CCE performed the statistical analyses. The content of this publication was determined by the authors. All authors contributed to the conception or design of the work, or the acquisition, analysis or interpretation of data. All authors reviewed and approved the manuscript. Funding: This study was funded by Arthritis Research UK (18810). Rituximab was provided free of charge by Hoffman La Roche. Disclaimer: The views expressed in this publication are those of the authors and not necessarily those of the NHS, the National Institute for Health Research or the Department of Health.

Contributors: BAF drafted the manuscript and CCE performed the statistical analyses. The content of this publication was determined by the authors. All authors contributed to the conception or design of the work, or the acquisition, analysis or interpretation of data. All authors reviewed and approved the manuscript. Funding: This study was funded by Arthritis Research UK (18810). Rituximab was provided free of charge by Hoffman La Roche. Disclaimer: The views expressed in this publication are those of the authors and not necessarily those of the NHS, the National Institute for Health Research or the Department of Health. Competing interests: BAF and SJB have received support from the NIHR Birmingham Biomedical Research Centre. BAF paid instructor/consultant for Novartis, Roche, BMS, Virtualscopics. W-FN is consultant for Pfizer, UCB, MedImmune, GSK, Takeda and Sanofi. MB is consultant for GSK, Amgen/MedImmune and UCB. SJB is consultant for Cellgene, Glenmark, GSK, Eli Lilly, Novartis, Roche, Takeda, UCB. Ethics approval: Leeds West Ethics Committee (ref. 10/H1307/99). Provenance and peer review: Not commissioned; externally peer reviewed. Correction notice: This article has been corrected since it published Online First. The affiliation for Anwar Tappuni has been corrected. Presented at: An abstract of this work was previously presented at the 2017 American College of Rheumatology Annual Congress.

Introduction Autoimmune connective tissue diseases (AI-CTDs) include systemic lupus erythematosus (SLE), primary Sjogren’s syndrome (pSS), systemic sclerosis, inflammatory myopathies, mixed and undifferentiated CTDs. A hallmark of their pathogenesis is loss of self-tolerance leading to autoreactivity and production of antibodies against self-nuclear antigens (ANAs). ANA can be detected in serum up to 10 years before clinical features, representing a phase of subclinical autoimmunity.1 However, ANA is present in up to 25% of the general population, of whom less than 1% develop clinical autoimmunity.2 3 Individuals with ANA therefore constitute At-Risk population of whom a minority will progress to AI-CTD.4 5 The factors that dictate whether this autoreactivity develops into autoimmune disease are unknown. But if these were understood and predictable, then effective intervention might be possible, preventing the severe disease and heavy glucocorticoid use for remission induction of a newly diagnosed AI-CTD.

ss to AI-CTD.4 5 The factors that dictate whether this autoreactivity develops into autoimmune disease are unknown. But if these were understood and predictable, then effective intervention might be possible, preventing the severe disease and heavy glucocorticoid use for remission induction of a newly diagnosed AI-CTD. Variants in type I interferon (IFN-I) pathway are prominent in the genetic susceptibility to AI-CTDs and therefore a focus for investigation.6–8 However, their role in disease initiation is currently unclear. IFN activity is usually quantified using expression of interferon-stimulated genes (ISGs). Interpretation of ISG expression is complex with multiple IFN subtypes produced by different cell types and tissues, as well as a transcriptional response in all nucleated cells with variation between cell types. Previously used IFN signatures have a categorical high/low classification9 10 or may have been affected by the ISGs selected.11–13 We recently described two continuous ISG expression scores (IFN-Score-A and IFN-Score-B) that in combination better identify clinically meaningful differences in IFN status between and within autoimmune diseases.14

tures have a categorical high/low classification9 10 or may have been affected by the ISGs selected.11–13 We recently described two continuous ISG expression scores (IFN-Score-A and IFN-Score-B) that in combination better identify clinically meaningful differences in IFN status between and within autoimmune diseases.14 In other autoimmune diseases such as rheumatoid arthritis (RA), early evidence of progression to disease may be found at a target tissue level.15 The tissues most commonly affected in AI-CTDs are the joints and skin. Musculoskeletal ultrasound can detect subclinical synovitis in SLE16 but has not been assessed in At-Risk individuals. In skin, specialised local immune processes are found in SLE. Previous studies comparing keratinocytes or skin biopsies isolated from patients with cutaneous lupus and healthy controls (HCs) found marked differences in IL-18R responsiveness,17 IFN-λ expression,18 as well as a role of IFN-κ in initiating a feed-forward loop, which promoted exaggerated ISG activation in cutaneous lupus.19 IFN-I status in the skin has not been assessed in At-Risk individuals. The aims of this study were to evaluate clinical, blood and tissue interferon and imaging biomarkers of progression from At-Risk to AI-CTD with a view to establish a strategy for disease prevention.

In other autoimmune diseases such as rheumatoid arthritis (RA), early evidence of progression to disease may be found at a target tissue level.15 The tissues most commonly affected in AI-CTDs are the joints and skin. Musculoskeletal ultrasound can detect subclinical synovitis in SLE16 but has not been assessed in At-Risk individuals. In skin, specialised local immune processes are found in SLE. Previous studies comparing keratinocytes or skin biopsies isolated from patients with cutaneous lupus and healthy controls (HCs) found marked differences in IL-18R responsiveness,17 IFN-λ expression,18 as well as a role of IFN-κ in initiating a feed-forward loop, which promoted exaggerated ISG activation in cutaneous lupus.19 IFN-I status in the skin has not been assessed in At-Risk individuals. The aims of this study were to evaluate clinical, blood and tissue interferon and imaging biomarkers of progression from At-Risk to AI-CTD with a view to establish a strategy for disease prevention. Methods Patients and design A prospective observational study was undertaken in individuals who were referred from primary care to Leeds Teaching Hospitals NHS Trust due to suspected AI-CTD between November 2014 and May 2017. Inclusion criteria were (1) ANA-positive of at least 1:80 titre on indirect immunofluorescence and using multiplex immunoassays (excluding those with scleroderma (centromere, Scl-70) or myositis-specific (PL-12, OJ, PL-7, Mi-2, Ku, Jo-1, PM-Scl75, PM-Scl100, SRP and EJ) antibodies only); (2) ≤1 clinical criterion based on 2012 Systemic Lupus International Collaborating Clinics Classification Criteria (SLICC)20 and not meeting classification criteria for other AI-CTD21–23 or RA24; (3) symptom duration <12 months; (4) glucocorticoid, antimalarial and immunosuppressive treatment-naïve. Forty-nine HCs and 114 patients with SLE were used as negative and positive controls.

al Collaborating Clinics Classification Criteria (SLICC)20 and not meeting classification criteria for other AI-CTD21–23 or RA24; (3) symptom duration <12 months; (4) glucocorticoid, antimalarial and immunosuppressive treatment-naïve. Forty-nine HCs and 114 patients with SLE were used as negative and positive controls. Assessment schedule and outcome Comprehensive assessments including clinical, laboratory, imaging, bloods and skin biomarkers were performed at baseline, 12 months and annually for 3 years. Participants were given a helpline number for an additional flare visit if they had new or worsening inflammatory symptoms. Progression was defined by meeting the 2012 SLICC criteria for SLE,20 2016 ACR/EULAR criteria for pSS21 or other relevant classification criteria for AI-CTD22 23 at 12 months as assessed by rheumatologists. Clinical and laboratory assessment Age, gender, ethnicity, history of first-degree or second-degree relative(s) with autoimmune rheumatic diseases (ARDs), smoking history, SLICC criteria for SLE,20 signs or glandular symptoms criteria for pSS,21 patient and physician global health assessment using 100 mm Visual Analogue Scale were recorded.

y assessment Age, gender, ethnicity, history of first-degree or second-degree relative(s) with autoimmune rheumatic diseases (ARDs), smoking history, SLICC criteria for SLE,20 signs or glandular symptoms criteria for pSS,21 patient and physician global health assessment using 100 mm Visual Analogue Scale were recorded. ANA was tested using indirect immunofluorescence and a panel of nuclear autoantibodies including anti-dsDNA, extractable nuclear antigens (including Ro52, Ro60, La, Sm, Chromatin, RNP, Sm/RNP and Ribosomal P) and antiphospholipid antibodies (Cardiolipin and B2-Glycoprotein IgGs) using Bioplex 2200 Immunoassay. Lupus anticoagulant tests including activated partial thromboplastin time (APTT) (Actin FS), APTT-synthetic phospholipid (with correction) and dilute Russell’s viper venom time (with correction) were deemed positive if persistent when repeated at 12 weeks. Full blood count was processed at a single accredited diagnostic laboratory. Complement levels (C3 and C4) were measured by nephelometry. Musculoskeletal ultrasound Ultrasound examination of wrists, metacarpophalangeal and proximal interphalangeal joints were performed by two rheumatologists, using General Electric S7 machine with a 6–15 MHz transducer. Outcome Measures in RA Clinical Trials (OMERACT) criteria25 were used to define synovitis, that is, the presence of grey-scale (GS) ≥grade 2 and/or power Doppler (PD) ≥grade 1.

angeal and proximal interphalangeal joints were performed by two rheumatologists, using General Electric S7 machine with a 6–15 MHz transducer. Outcome Measures in RA Clinical Trials (OMERACT) criteria25 were used to define synovitis, that is, the presence of grey-scale (GS) ≥grade 2 and/or power Doppler (PD) ≥grade 1. Blood and skin IFN scores A two-score system of ISGs, as previously described,14 was calculated without the knowledge of participant’s clinical status. See online supplementary file for details. Briefly, peripheral blood mononuclear cells (PBMCs) were separated using density gradient method (Lymphoprep; Alere Technologies, Norway) from EDTA-anticoagulated blood. Total RNA purification kit (Norgen Biotek, Canada) was used followed by quantitative real-time reverse transcriptase-PCR (qRT-PCR) using TaqMan assays (Applied Biosystems, Invitrogen) for the selected 30 ISGs.7 These assays were performed using the BioMark HD System with appropriate cycling protocols for the 96.96 chip. Data were normalised using Peptidylprolyl isomerase A as a reference gene to calculate ΔCt. 10.1136/annrheumdis-2018-213386.supp4Supplementary data Factor analysis was used to reduce the 30 ISGs into a smaller number of factors.26 Two factors, IFN-Score-A and IFN-Score-B, explained 84% of the variance with limited cross-loading. Factor scores were calculated as the median level of expression of the genes loaded by each factor.

10.1136/annrheumdis-2018-213386.supp4Supplementary data Factor analysis was used to reduce the 30 ISGs into a smaller number of factors.26 Two factors, IFN-Score-A and IFN-Score-B, explained 84% of the variance with limited cross-loading. Factor scores were calculated as the median level of expression of the genes loaded by each factor. Skin biopsy One 4 mm biopsy was obtained from non-lesional non-sun-exposed areas (upper back or upper arms) of At-Risk individuals (n=10) and HCs (n=6), and from active lesions of patients with SLE (n=10). Biopsies were snap frozen in optimum cutting temperature (OCT) compound and sectioned at a thickness of 5 µm ensuring no remaining OCT material contaminating subsequent RNA extraction/RT procedures. Gene expression analysis and calculation of factor scores were conducted as for PBMCs.

sions of patients with SLE (n=10). Biopsies were snap frozen in optimum cutting temperature (OCT) compound and sectioned at a thickness of 5 µm ensuring no remaining OCT material contaminating subsequent RNA extraction/RT procedures. Gene expression analysis and calculation of factor scores were conducted as for PBMCs. Statistical analyses Associations between categorical variables were tested by Fisher’s exact and Stuart-Maxwell tests for independent and paired samples, respectively. Continuous variables were compared using either Student’s t-tests or analysis of variance (ANOVA) followed by pairwise Tukey tests. For associations, Kendall’s tau-b correlation was used if ties were present, otherwise using Pearson’s correlation. Receiver operator curves (ROCs) were used to assess predictive strength and identify optimal thresholds for predicting progression to AI-CTD. For 13 At-Risk patients, gene expression data were missing at random due to samples not being processed on the day. For comparisons with HC and SLE groups, only At-Risk patients with complete data were presented. For prediction of progression, multiple imputation by chained equations was used to create 20 complete datasets, results of which were combined according to Rubin’s rules. Multivariable analyses were performed using penalised logistic regression by Lasso method.27 Leave-one-out cross-validation (R package cv.glmnet)28 identified the largest penalty coefficient lambda within 1 SE of the value that minimised deviance in each imputed dataset; average coefficients from the best models were calculated. All analyses of IFN Scores were conducted using ∆Ct scaling; results were then converted to relative expression (2−ΔCt) or fold difference (FD) (2−ΔΔCt).

argest penalty coefficient lambda within 1 SE of the value that minimised deviance in each imputed dataset; average coefficients from the best models were calculated. All analyses of IFN Scores were conducted using ∆Ct scaling; results were then converted to relative expression (2−ΔCt) or fold difference (FD) (2−ΔΔCt). Statistical analyses were performed using Stata V.13.1 (StataCorp, College Station, Texas, USA), R V.3.3.329 and GraphPad Prism V.7.03 (GraphPad, La Jolla, California, USA) for Windows. Results Patient characteristics The flowchart of participants is presented in figure 1. A total of 135 At-Risk individuals were recruited. Of these, 118 had at least 12 months of follow-up and were analysed. Baseline characteristics are described in table 1. Figure 1 Flowchart of the At-Risk study in Leeds. AI-CTD, autoimmune-related connective tissue disease; ANA, antinuclear antibody; SLE, systemic lupus erythematosus. Table 1 Baseline characteristics of the 118 At-Risk of AI-CTD individuals

Results Patient characteristics The flowchart of participants is presented in figure 1. A total of 135 At-Risk individuals were recruited. Of these, 118 had at least 12 months of follow-up and were analysed. Baseline characteristics are described in table 1. Figure 1 Flowchart of the At-Risk study in Leeds. AI-CTD, autoimmune-related connective tissue disease; ANA, antinuclear antibody; SLE, systemic lupus erythematosus. Table 1 Baseline characteristics of the 118 At-Risk of AI-CTD individuals Age, median (range) years 48 (20–84) No of female patients (%) 104 (88) Ethnicity, N (%) Caucasian 85 (72) Indian/South Asian 20 (17) African/Caribbean 12 (10) Chinese 1 (1) Positive ANA, N (%) 118 (100) No of positive ANA specificities, median (range) 1 (1–4) Autoantibody-positive specificities, N (%) Anti-dsDNA 42 (36) 10–20 IU/mL 15 (13) 21–50 IU/mL 18 (15) >50 IU/mL 9 (8) Anti-Ro 50 (42) <8 AI 24 (20) ≥8 AI 26 (22) Anti-La 9 (8) Anti-Smith 5 (4) Anti-Chromatin 17 (14) Anti-RNP 2 (2) Anti-Ribosomal P 0 (0) Anti-Sm/RNP 16 (14) Anti-Cardiolipin/anti-B2-glycoprotein 5 (4) Positive lupus anticoagulant, N (%) 4 (3) Concurrent positive RF, N (%) 11 (9) Low titre (<50 iU/mL), N (%) 5 (4) High titre (≥50 iU/mL), N (%) 6 (5) Concurrent positive anti-CCP antibody, N (%)* 3 (3) Low complement levels (C3 or C4), N (%) 8 (7) No of clinical criteria, N (%) 0 20 (17) 1 98 (83) Clinical criteria present, N (%) Acute or sub-acute cutaneous lupus erythematosus† 27 (24) Chronic cutaneous lupus erythematosus 1 (1) Oral or nasal ulcers 4 (3) Non-scarring alopecia 5 (4) Arthritis 43 (36) Serositis 1 (1) Renal 0 (0) Neurological 0 (0) Haemolytic anaemia 0 (0) Leucopaenia or lymphopaenia 12 (10) Thrombocytopenia 5 (4) Glandular signs 0 (0) Family history of autoimmune rheumatic disease, N (%)‡ 43 (36) Ever smoked, N (%) 45 (38) *All patients had low anti-CCP antibody titre (<50 U/mL).

) Arthritis 43 (36) Serositis 1 (1) Renal 0 (0) Neurological 0 (0) Haemolytic anaemia 0 (0) Leucopaenia or lymphopaenia 12 (10) Thrombocytopenia 5 (4) Glandular signs 0 (0) Family history of autoimmune rheumatic disease, N (%)‡ 43 (36) Ever smoked, N (%) 45 (38) *All patients had low anti-CCP antibody titre (<50 U/mL). †Only 1 patient had SCLE lesion. ‡First-degree or second-degree relative with autoimmune rheumatic disease. AI-CTD, autoimmune-related connective tissue disease; ANA, antinuclear antibody; CCP, cyclic citrullinated peptide; dsDNA, double-stranded DNA; RF, rheumatoid factor; RNP, ribonucleic protein. Clinical outcomes at 12 months At 12 months, 19/118 (16 %) At-Risk individuals progressed to a diagnosis of AI-CTD. These were SLE (n=14; 74%) and pSS (n=5; 26%). In those who progressed, all had one clinical criterion at baseline. The number of clinical SLE criteria increased to 2 in 4/19 (21%), 3 in 9/19 (47%) and 4 in 6/19 (32%) (Stuart-Maxwell χ2=20.0, p<0.001) at 12 months. These details are presented in table 2 and online supplementary figure S1. Two patients developed internal organ involvement; pleural effusion and class III lupus nephritis. 10.1136/annrheumdis-2018-213386.supp1Supplementary data Table 2 Clinical characteristics of At-Risk progressors at 12 months

Clinical outcomes at 12 months At 12 months, 19/118 (16 %) At-Risk individuals progressed to a diagnosis of AI-CTD. These were SLE (n=14; 74%) and pSS (n=5; 26%). In those who progressed, all had one clinical criterion at baseline. The number of clinical SLE criteria increased to 2 in 4/19 (21%), 3 in 9/19 (47%) and 4 in 6/19 (32%) (Stuart-Maxwell χ2=20.0, p<0.001) at 12 months. These details are presented in table 2 and online supplementary figure S1. Two patients developed internal organ involvement; pleural effusion and class III lupus nephritis. 10.1136/annrheumdis-2018-213386.supp1Supplementary data Table 2 Clinical characteristics of At-Risk progressors at 12 months Clinical criteria Baseline 12 months (n=19) (n=19) Mucocutaneous ACLE or SCLE 5/19 (26%) 13/19 (68%) Mucosal ulcers 2/19 (11%) 8/19 (42%) Alopecia 0 4/19 (21%) Musculoskeletal Synovitis 9/19 (47%) 18/19 (95%) Haematological Leucopaenia or lymphopenia 3/19 (16%) 7/19 (37%) Thrombocytopenia 0 1/19 (5%) Glandular signs 0 6/19 (32%) Serositis Pleural effusion 0 1/19 (5%) Renal Class III nephritis 0 1/19 (5%) ACLE, acute cutaneous lupus erythematosus; SCLE, sub-acute cutaneous lupus erythematosus. In contrast, 19/99 (19%) of the non-progressors had no clinical SLE criteria at both baseline and 12 months, 1/99 (1%) increased from 0 to 1, 41/99 (42%) decreased from 1 to 0 indicating a remission of autoimmunity and 38/99 (38%) had one criterion at both time points (Stuart-Maxwell χ2=38.1, p<0.001).

Clinical criteria Baseline 12 months (n=19) (n=19) Mucocutaneous ACLE or SCLE 5/19 (26%) 13/19 (68%) Mucosal ulcers 2/19 (11%) 8/19 (42%) Alopecia 0 4/19 (21%) Musculoskeletal Synovitis 9/19 (47%) 18/19 (95%) Haematological Leucopaenia or lymphopenia 3/19 (16%) 7/19 (37%) Thrombocytopenia 0 1/19 (5%) Glandular signs 0 6/19 (32%) Serositis Pleural effusion 0 1/19 (5%) Renal Class III nephritis 0 1/19 (5%) ACLE, acute cutaneous lupus erythematosus; SCLE, sub-acute cutaneous lupus erythematosus. In contrast, 19/99 (19%) of the non-progressors had no clinical SLE criteria at both baseline and 12 months, 1/99 (1%) increased from 0 to 1, 41/99 (42%) decreased from 1 to 0 indicating a remission of autoimmunity and 38/99 (38%) had one criterion at both time points (Stuart-Maxwell χ2=38.1, p<0.001). Notably, 1/99 (1%) of non-progressors had ankylosing spondylitis while 4/99 (4%) of had cancers (lung=1, hepatocellular=1, prostate=1 and leiomyosarcoma=1). Interferon status in At-Risk differs from SLE At baseline, IFN-Score-A differed between groups (ANOVA F=40.26; p<0.001). It was increased relative to HC (n=49) in both At-Risk (n=105; FD (95% CI) 2.21 (1.22 to 4.00), p=0.005) and SLE (n=114; 7.81 (4.33 to 14.04), p<0.001), and was increased in SLE relative to At-Risk (3.54 (2.22 to 5.63), p<0.001) (figure 2A). In contrast, although IFN-Score-B differed between groups overall (F=63.35; p<0.001), it did not differ between At-Risk and HC (0.98 (0.66 to 1.46), p=0.993), but was increased in SLE to both HC (3.85 (2.60 to 5.72), p<0.001) and At-Risk (3.93 (2.87 to 5.37), p<0.001) (figure 2B).

(2.22 to 5.63), p<0.001) (figure 2A). In contrast, although IFN-Score-B differed between groups overall (F=63.35; p<0.001), it did not differ between At-Risk and HC (0.98 (0.66 to 1.46), p=0.993), but was increased in SLE to both HC (3.85 (2.60 to 5.72), p<0.001) and At-Risk (3.93 (2.87 to 5.37), p<0.001) (figure 2B). Figure 2 Pattern of baseline interferon scores and their relationships with clinical immunology markers. (A) Baseline expression of IFN-Score-A was higher in At-Risk individuals compared with healthy controls. (B) However, there was no difference in IFN-Score-B between both groups. ***Highly significant (p<0.001), **moderate significant (0.001<p<0.01), *significant (0.01<p<0.05). (C, D) Both IFN scores were not correlated with the number of positive antinuclear antibody (ANA) specificities (ie, sum of anti-dsDNA, Ro, La, Sm, Chromatin, RNP, Sm/RNP and Ribosomal P) and (E, F) there were only weak correlations between IFN-Score-A and complement and lymphocyte count. Data for gene expression were expressed as reflected values for ∆Ct so that higher IFN Scores represented greater expression. Relationships of interferon scores with autoantibodies, complement and lymphopaenia Correlations between routine immunology markers and IFN Scores were performed in observed data using reflected ∆Ct so that higher IFN Scores represented greater expression. At baseline, there was no association between number of positive ANA specificities (ie, anti-dsDNA, Ro, RNP etc.) and IFN-Score-A (n=105, Kendall’s tau-b 0.13, p=0.084) or IFN-Score-B (tau-b 0.09, p=0.234) (figure 2C,D).

rmed in observed data using reflected ∆Ct so that higher IFN Scores represented greater expression. At baseline, there was no association between number of positive ANA specificities (ie, anti-dsDNA, Ro, RNP etc.) and IFN-Score-A (n=105, Kendall’s tau-b 0.13, p=0.084) or IFN-Score-B (tau-b 0.09, p=0.234) (figure 2C,D). The titres of two antibodies that were mostly prevalent using Bioplex, anti-dsDNA and anti-Ro, were divided into three and two groups, respectively. There were no differences in both IFN Scores among the three anti-dsDNA groups (online supplementary figure S2A,B). Elevated levels of IFN-Score-A (FD 2.41 (95% CI 1.10 to 5.26)) but not Score-B were found in the high titre, that is, ≥8 AI anti-Ro antibody positive group (online supplementary figure S2C, D). 10.1136/annrheumdis-2018-213386.supp2Supplementary data There was a weak negative correlation between C4 levels and IFN-Score-A (n=97, Pearson’s r=−0.221, p=0.029) (figure 2E) but not IFN-Score-B (r=−0.089, p=0.385). There was a weak negative correlation between lymphocyte count and IFN-Score-A (n=105, r=−0.230, p=0.018) (figure 2F) but not IFN-Score-B (r=−0.127, p=0.195). Baseline interferon status in skin In parallel to results obtained for PBMC, at baseline only IFN-Score-A was increased in non-lesional skin biopsies in At-Risk (n=10) versus HC (n=6); FD 28.74 (1.29 to 639.48), p=0.036. There was no difference in IFN-Score-B; FD 1.82 (0.86 to 3.86), p=0.100. As expected, both IFN Scores were higher in SLE (active lesions) compared with either At-Risk or HC; all p<0.05.

ne only IFN-Score-A was increased in non-lesional skin biopsies in At-Risk (n=10) versus HC (n=6); FD 28.74 (1.29 to 639.48), p=0.036. There was no difference in IFN-Score-B; FD 1.82 (0.86 to 3.86), p=0.100. As expected, both IFN Scores were higher in SLE (active lesions) compared with either At-Risk or HC; all p<0.05. Comparison of baseline interferon status between blood and skin Expression of both IFN Scores was higher in At-Risk versus HC in both skin and PBMC, but FDs were greater in skin (figure 3C). This might have been due to the small sample size for skin samples (paired skin–PBMC samples were not available). Figure 3 Baseline interferon (IFN) scores in bloods as prognostic biomarkers. (A–B) Baseline expression of both IFN-Score-A and IFN-Score-B were higher in At-Risk individuals who progressed to autoimmune-related connective tissue disease compared with the non-progressors, but to a greater fold difference in the latter. ***Highly significant (p<0.001), **moderately significant (0.001<p<0.01), *significant (0.01<p<0.05). (C) Fold differences for both IFN scores between At-Risk and healthy controls (HCs) were greater in skin than bloods. (D) The area under the receiver operating characteristic curve was significantly greater for IFN-Score-B than IFN-Score-A. The blue arrow denotes the optimal cut-off using Youden’s index while the red arrow denotes the proposed cut-off for prevention study. PBMC, peripheral blood mononuclear cell; SLE, systemic lupus erythematosus.

under the receiver operating characteristic curve was significantly greater for IFN-Score-B than IFN-Score-A. The blue arrow denotes the optimal cut-off using Youden’s index while the red arrow denotes the proposed cut-off for prevention study. PBMC, peripheral blood mononuclear cell; SLE, systemic lupus erythematosus. Prediction of AI-CTD using baseline interferon scores in blood When At-Risk were divided according to AI-CTD progression status at 12 months, both IFN Scores differed among the groups overall (p<0.001) and both were elevated in At-Risk progressors (n=19) versus non-progressors (n=86), to a greater extent for IFN-Score-B (FD 3.22 (1.74 to 5.95), p<0.001) than IFN-Score-A (2.94 (1.14, 7.54), p=0.018) (figure 3A,B). Non-progressors did not differ from HC (n=49) for both scores; IFN-Score-B (0.79 (0.51 to 1.23), p=0.520) and IFN-Score-A (1.82 (0.93 to 3.53), p=0.096). Neither IFN Score differed between At-Risk progressors and SLE (both p>0.1). Since the number of skin biopsies obtained in At-Risk was small (n=10), no formal association between IFN Scores and progression could be determined.

Prediction of AI-CTD using baseline interferon scores in blood When At-Risk were divided according to AI-CTD progression status at 12 months, both IFN Scores differed among the groups overall (p<0.001) and both were elevated in At-Risk progressors (n=19) versus non-progressors (n=86), to a greater extent for IFN-Score-B (FD 3.22 (1.74 to 5.95), p<0.001) than IFN-Score-A (2.94 (1.14, 7.54), p=0.018) (figure 3A,B). Non-progressors did not differ from HC (n=49) for both scores; IFN-Score-B (0.79 (0.51 to 1.23), p=0.520) and IFN-Score-A (1.82 (0.93 to 3.53), p=0.096). Neither IFN Score differed between At-Risk progressors and SLE (both p>0.1). Since the number of skin biopsies obtained in At-Risk was small (n=10), no formal association between IFN Scores and progression could be determined. Baseline IFN-Score-B threshold of progression to AI-CTD Prognostic ability of baseline IFN Scores to predict progression to AI-CTD at 12 months was assessed using ROC curve analysis. The area under the ROC curve was greater for IFN-Score-B (0.82 (95% CI 0.73 to 0.92)) than IFN-Score-A (0.70 (0.57 to 0.83)); χ2=4.19, p=0.041. A cut-off of ≤5.01 ∆Ct for IFN-Score-B maximised the Youden’s index (sensitivity+specificity−1) yielding 95% (95% CI 75% to 99%) sensitivity, 60% (50% to 70%) specificity, 35% (23% to 48%) positive predictive value (PPV) and 98% (90% to >99%) negative predictive value (NPV). However, for a rule-in biomarker for future prevention studies, a high specificity is required to exclude individuals with the lowest risk. For this purpose, we propose a cut-off of ≤3.90 ∆Ct that resulted in 68% (46% to 85%) sensitivity, 80% (70% to 88%) specificity, 43% (27% to 61%) PPV and 92% (84% to 96%) NPV (figure 3D).

, for a rule-in biomarker for future prevention studies, a high specificity is required to exclude individuals with the lowest risk. For this purpose, we propose a cut-off of ≤3.90 ∆Ct that resulted in 68% (46% to 85%) sensitivity, 80% (70% to 88%) specificity, 43% (27% to 61%) PPV and 92% (84% to 96%) NPV (figure 3D). Baseline IFN Scores were lower in At-Risk without versus with one clinical criterion All 20/118 (17%) At-Risk individuals who had no SLE clinical criterion at baseline did not progress to AI-CTD at 12 months. At baseline, FDs for both IFN scores differed among the groups overall (p<0.001) and both were lower in At-Risk with no criterion (n=17) versus with one criterion (n=88); all p<0.05 (online supplementary figure S3 in the online supplementary file). 10.1136/annrheumdis-2018-213386.supp3Supplementary data Musculoskeletal ultrasound Of the 117 At-Risk individuals with ultrasound available, 21 (18%) had ultrasound-defined synovitis at baseline (GS ≥2 only=13, PD ≥1 with or without GS ≥2=8). Of the 20 individuals who progressed, 7 (35%) had positive ultrasound at baseline versus 14% of non-progressors; p=0.050, PPV (95% CI)=33% (17% to 55%), NPV 86% (78% to 92%).

Risk individuals with ultrasound available, 21 (18%) had ultrasound-defined synovitis at baseline (GS ≥2 only=13, PD ≥1 with or without GS ≥2=8). Of the 20 individuals who progressed, 7 (35%) had positive ultrasound at baseline versus 14% of non-progressors; p=0.050, PPV (95% CI)=33% (17% to 55%), NPV 86% (78% to 92%). Furthermore, 43/118 of At-Risk individuals had clinical arthritis based on SLICC20 (8/43 (19%) had ≥2 joints with swelling or effusion while 35/43 (81%) had ≥2 joints with tenderness and early morning stiffness of ≥30 min) while 75/118 had no arthritis. In those without arthritis, ultrasound-defined synovitis was detected in 10/75 (13%) and 4/10 (40%) progressed to AI-CTD. Conversely, in those with arthritis, only 11/42 (26%) had ultrasound-defined synovitis and 3/11 (27%) progressed to AI-CTD at 12 months. Sensitivity and specificity of physician-judged arthritis with ultrasound-defined synovitis were 52% and 68%, respectively.

etected in 10/75 (13%) and 4/10 (40%) progressed to AI-CTD. Conversely, in those with arthritis, only 11/42 (26%) had ultrasound-defined synovitis and 3/11 (27%) progressed to AI-CTD at 12 months. Sensitivity and specificity of physician-judged arthritis with ultrasound-defined synovitis were 52% and 68%, respectively. Multivariable analysis of baseline predictors of progression to AI-CTD In imputed univariable analyses, all putative predictors were associated with progression to AI-CTD at 12 months at the 10% level of significance except for complement level and lymphocyte count (both p>0.1), which were excluded from multivariable analysis (table 3). In multivariable logistic regression, family history of ARDs (OR 8.20, p=0.012) and IFN-Score-B (OR=3.79, p=0.005) were independently associated with progression. Penalised ORs remained substantive for these variables when all other variables were removed from the model. Results in complete data (n=100) were similar (data not shown). Table 3 Penalised logistic regression for predictors of progression to autoimmune-related connective tissue disease at 12 months

Multivariable analysis of baseline predictors of progression to AI-CTD In imputed univariable analyses, all putative predictors were associated with progression to AI-CTD at 12 months at the 10% level of significance except for complement level and lymphocyte count (both p>0.1), which were excluded from multivariable analysis (table 3). In multivariable logistic regression, family history of ARDs (OR 8.20, p=0.012) and IFN-Score-B (OR=3.79, p=0.005) were independently associated with progression. Penalised ORs remained substantive for these variables when all other variables were removed from the model. Results in complete data (n=100) were similar (data not shown). Table 3 Penalised logistic regression for predictors of progression to autoimmune-related connective tissue disease at 12 months Baseline predictors No progression n=99 Progression n=19 Univariable OR (95% CI), p values Multivariable OR (95% CI), p values Penalised coefficient to OR Age, mean (SD) 49.0 (15.8) 39.6 (11.9) 0.96 (0.93 to 0.99), 0.016 0.97 (0.92 to 1.02), 0.232 0.000 to 1.000 Ever smoked, (%) 41.8% 20.0% 0.35 (0.11 to 1.12), 0.076 0.34 (0.06 to 1.91), 0.222 0.000 to 1.000 Family history of ARDs (%) 30.6% 65.0% 4.21 (1.53 to 11.61), 0.005 8.20 (1.58 to 42.53), 0.012 0.243 to 1.275 No of positive ANA specificities, median (IQR) 1 (1–1) 1 (1–2) 2.07 (0.97 to 4.40), 0.060 2.41 (0.71 to 8.20), 0.161 0.000 to 1.000 Complement C4 level, mean (SD) 0.29 (0.12) 0.26 (0.08) 0.06 (0.00 to 8.05), 0.264 Excluded Excluded Lymphocyte count, mean (SD) 2.04 (0.77) 1.83 (0.67) 0.67 (0.34 to 1.34), 0.257 Excluded Excluded No of joints with positive ultrasound for synovitis, median (IQR) 0 (0–0) 0 (0–2) 1.20 (0.97 to 1.47), 0.086 1.44 (0.98 to 2.11), 0.061 0.002 to 1.002 Patient VAS, median (IQR) 36 (16–61) 47 (26–75) 1.02 (1.00 to 1.04), 0.079 1.01 (0.98 to 1.04), 0.484 0.000 to 1.000 Physician VAS, median (IQR) 11 (3–31) 31 (15–47) 1.04 (1.01 to 1.06), 0.008 1.01 (0.97 to 1.06), 0.618 0.000 to 1.000 IFN-Score-A (−ΔCt), mean (SD)* −5.3 (1.9) −3.8 (2.26) 1.43 (1.11 to 1.84), 0.005 0.87 (0.54 to 1.39), 0.560 0.000 to 1.000 IFN-Score-B (−ΔCt), mean (SD)* −5.3 (1.4) −3.7 (1.0) 2.55 (1.60 to 4.08), <0.001 3.79 (1.50 to 9.58), 0.005 0.319 to 1.376 *Analysis was made based on reflected ∆Ct. Thus, the higher the number, the higher the gene expression to give positive values for ORs.

.11 to 1.84), 0.005 0.87 (0.54 to 1.39), 0.560 0.000 to 1.000 IFN-Score-B (−ΔCt), mean (SD)* −5.3 (1.4) −3.7 (1.0) 2.55 (1.60 to 4.08), <0.001 3.79 (1.50 to 9.58), 0.005 0.319 to 1.376 *Analysis was made based on reflected ∆Ct. Thus, the higher the number, the higher the gene expression to give positive values for ORs. ANA, antinuclear antibody; ARD, autoimmune rheumatic disease; IFN, interferon; VAS, Visual Analogue Score. Discussion In this study, we report a unique cohort of At-Risk of AI-CTD individuals with longitudinal follow-up until progression to clinical autoimmunity. We demonstrate that IFN activity is strongly associated with progression independent of baseline clinical status, with measurement according to a two-score system we described being crucial. These results provide a rationale for diagnostic and preventative treatment pathways as well as assert the importance of interferons in disease initiation. Referrals of ANA-positive individuals to rheumatologists has increased over the last decade.30 Concerns are that these At-Risk individuals may be discharged prematurely or be observed in an inefficient ‘watch and wait’ fashion until the diagnosis is clear, by which time the potential to prevent disease and confer the most benefit may be lost. Thus, by undertaking the largest prospective study of At-Risk individuals, which is the first to integrate clinical, imaging and immunological assessments (including skin), our findings offer a novel approach, biomarkers and have implications for future development of targeted therapies for this group of patients.

ost. Thus, by undertaking the largest prospective study of At-Risk individuals, which is the first to integrate clinical, imaging and immunological assessments (including skin), our findings offer a novel approach, biomarkers and have implications for future development of targeted therapies for this group of patients. Within ANA-positive individuals, different immune phenotypes could be defined. At baseline, IFN-Score-A was elevated but not IFN-Score-B compared with HC. However, IFN-Score-B (and to a lesser degree, IFN-Score-A) were mostly elevated in those who progressed to AI-CTD. IFN-Score-A comprises many well-known ISGs that respond to IFN-I (IFN-α, IFN-β, IFN-κ, IFN-ω). In contrast, IFN-Score-B comprises ISGs that coincide with M3.4 and M5.12 modules of a previous microarray study.7 These ISGs were suggested to be responsive to IFN-II (IFN-γ), IFN-III (IFN-λ) as well as IFN-I. However, we cannot exclude the influence of other inflammatory mediators on this pattern of gene expression.14 Some studies suggested that IFN-I contributes to priming cells to secrete IFN-II.31 32 Conversely, a study that measured IFN activity from serum postulated a sequential role of IFN-II augmentation that led to autoantibody accumulation and subsequent elevations in IFN-α prior to SLE.33 Although we could not confirm which IFN pathways predominate, our findings suggest that progression to AI-CTD may not be exclusively driven by IFN-I but by a synergistic activation of ISGs induced by a range of IFNs and IFN-Score-B could act as a biomarker for more diverse immune activation.

ns in IFN-α prior to SLE.33 Although we could not confirm which IFN pathways predominate, our findings suggest that progression to AI-CTD may not be exclusively driven by IFN-I but by a synergistic activation of ISGs induced by a range of IFNs and IFN-Score-B could act as a biomarker for more diverse immune activation. At the tissue level, this is the first study that quantifies IFN activity in non-lesional skin of At-Risk individuals. Interestingly, similar patterns of immune dysregulation were shown between skin and PBMC. However, markedly greater FDs in both IFN scores were found in the former compared with the latter, thus highlighting skin as a potential site of AI-CTD initiation. Only a third of the At-Risk individuals who had ultrasound-defined synovitis progressed to AI-CTD within 12 months. Additionally, small numbers of asymptomatic patients with ultrasound-detected synovitis were identified, so further work is required to determine the role of ultrasound in assessing At-Risk individuals. Together with a family history of ARD, IFN-Score-B from blood is independently predictive of progression and is convenient as a biomarker. We have defined a cut-off level of IFN-Score-B with a moderate diagnostic accuracy in order to design a prevention study.

Only a third of the At-Risk individuals who had ultrasound-defined synovitis progressed to AI-CTD within 12 months. Additionally, small numbers of asymptomatic patients with ultrasound-detected synovitis were identified, so further work is required to determine the role of ultrasound in assessing At-Risk individuals. Together with a family history of ARD, IFN-Score-B from blood is independently predictive of progression and is convenient as a biomarker. We have defined a cut-off level of IFN-Score-B with a moderate diagnostic accuracy in order to design a prevention study. This study has some limitations. First, the cohort was recruited from secondary care as well as positive ANA detected by both Bioplex and indirect immunofluorescence, which might contribute to moderate-to-high pre-test probabilities for AI-CTD. Thus, our results might not be generalised to all ANA-positive cases in primary care setting. However, our cohort was quite heterogenous in terms of ethnicity and 17% of the patients had no SLE criterion at baseline. Second, we excluded individuals with scleroderma or myositis-specific only autoantibodies, which might lead to preponderance of progression to SLE or pSS. Surprisingly, one patient had a severe ankylosing spondylitis and required biological therapy. Moreover, 4% of non-progressors had cancers thus highlighting the need to be vigilant of paraneoplastic manifestation in ANA-positive individuals as well as diverse alternative diagnoses in general. Lastly, sample size was still relatively small for multivariable analysis. However, we used penalised logistic regression to minimise overfitting of data.

ad cancers thus highlighting the need to be vigilant of paraneoplastic manifestation in ANA-positive individuals as well as diverse alternative diagnoses in general. Lastly, sample size was still relatively small for multivariable analysis. However, we used penalised logistic regression to minimise overfitting of data. In conclusion, a novel ISG score, IFN-Score-B and family history of ARD predict progression from ANA positivity to AI-CTD. After validation, the predictive value of IFN scores may allow us to identify patients with imminent AI-CTD for earlier intervention using therapies that block IFNs or conventional immunosuppressants to avoid irreversible organ damage and glucocorticoid exposure. Additionally, patients with benign autoreactivity can be better identified. The authors would like to thank the clinicians, study coordinators and laboratory technicians at the Leeds Pre-Connective Tissue Disease Clinic particularly Maya Buch, Sinisa Savic, Ai Lyn Tan, Francesco Del Galdo, Jacqueline Nam, Khaled Mahmoud, Huma Cassamoali, Sabina Khan, Diane Corscadden, Katie Mbara and Zoe Wigston for their substantial contribution in the acquisition of the data. MYMY and AP contributed equally. Handling editor: Josef S Smolen

The authors would like to thank the clinicians, study coordinators and laboratory technicians at the Leeds Pre-Connective Tissue Disease Clinic particularly Maya Buch, Sinisa Savic, Ai Lyn Tan, Francesco Del Galdo, Jacqueline Nam, Khaled Mahmoud, Huma Cassamoali, Sabina Khan, Diane Corscadden, Katie Mbara and Zoe Wigston for their substantial contribution in the acquisition of the data. MYMY and AP contributed equally. Handling editor: Josef S Smolen Contributors: MYMY, AP and EMV: substantial contributions to the conception or design of the work, or the acquisition, analysis or interpretation of data, drafting the work or revising it critically for important intellectual content, final approval of the version published and agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. YME-S, EMAH, KD, SU-H, MS, AA, MW and PE: substantial contributions to the conception or design of the work, or the acquisition, analysis or interpretation of data, drafting the work or revising it critically for important intellectual content and final approval of the version published. Funding: This research was funded/supported by the National Institute for Health Research (NIHR) and NIHR Leeds Biomedical Research Centre based at Leeds Teaching Hospitals NHS Trust (DRF-2014-07-155 and CS-2013-13-032). Disclaimer: The views expressed are those of the authors and not necessarily those of the NHS, the NIHR or the Department of Health.

Funding: This research was funded/supported by the National Institute for Health Research (NIHR) and NIHR Leeds Biomedical Research Centre based at Leeds Teaching Hospitals NHS Trust (DRF-2014-07-155 and CS-2013-13-032). Disclaimer: The views expressed are those of the authors and not necessarily those of the NHS, the NIHR or the Department of Health. Competing interests: EMV is an NIHR Clinician Scientist. He has received honoraria and research grant support from Roche, GSK and AstraZeneca. PE has received consultant fees from BMS, Abbott, Pfizer, MSD, Novartis, Roche and UCB. He has received research grants paid to his employer from Abbott, BMS, Pfizer, MSD and Roche. MW has received honoraria for educational activity and consultancy from Novartis, Janssen, Abbvie and Cellgene. ASZ has received honoraria from Roche/Chugai, BMS, Biogen and Menarini. Patient consent: Not required. Ethics approval: All individuals provided informed written consent and this research was carried out in compliance with the Declaration of Helsinki. The patients’ blood samples used for this study were collected under ethical approval, REC 10/H1306/88, National Research Ethics Committee Yorkshire and Humber–Leeds East, and healthy control participants’ peripheral blood was collected under the study number 04/Q1206/107. All experiments were performed in accordance with relevant guidelines and regulations. The University of Leeds was contracted with administrative sponsorship. Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: None.

Ethics approval: All individuals provided informed written consent and this research was carried out in compliance with the Declaration of Helsinki. The patients’ blood samples used for this study were collected under ethical approval, REC 10/H1306/88, National Research Ethics Committee Yorkshire and Humber–Leeds East, and healthy control participants’ peripheral blood was collected under the study number 04/Q1206/107. All experiments were performed in accordance with relevant guidelines and regulations. The University of Leeds was contracted with administrative sponsorship. Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: None. Correction notice: This article has been corrected since it published Online First. The affiliations have been updated.

Introduction Systemic sclerosis (SSc) or scleroderma is an autoimmune connective tissue disease characterised by accumulation of extracellular matrix (ECM) proteins within the affected tissues and a widespread vasculopathy comprising both defective angiogenesis and fibroproliferative vasculopathy.1–4 It is well established that transforming growth factor beta (TGF-β) plays a key role in accumulation of collagen and ECM proteins and downregulation of caveolin-1 (reviewed in Del Galdo et al 5and Lafyatis6). Nevertheless, the pathogenesis of sustained microangiopathy and defective angiogenesis, and their causal links with tissue fibrosis are less clearly understood. Recent studies employing a mouse strain with a ligand-dependent upregulation of TGF-β signalling in tissue fibroblasts (TBRIIΔk-fib) demonstrate that increased TGF-β signalling favours the onset of fibroproliferative vasculopathy typical of SSc following minimal endothelial cell injury,7 suggesting that TGF-β activation may also play a role in the pathogenesis of vasculopathy in SSc.

regulation of TGF-β signalling in tissue fibroblasts (TBRIIΔk-fib) demonstrate that increased TGF-β signalling favours the onset of fibroproliferative vasculopathy typical of SSc following minimal endothelial cell injury,7 suggesting that TGF-β activation may also play a role in the pathogenesis of vasculopathy in SSc. Proteomic studies have previously identified pigment epithelial-derived factor (PEDF) as one of the most abundant secreted proteins by SSc skin fibroblasts compared with healthy controls.8 PEDF is a 46 kDa secreted glycoprotein that belongs to the serpin superfamily but has no protease inhibitory function. Despite the lack of protease activity, PEDF exerts diverse physiological functions including antiangiogenesis,9 antivasopermeability10 and neurotrophic activities.11 12 PEDF is expressed abundantly in pigmented epithelium of the cornea where it plays a crucial role in the suppression of angiogenesis.9 13–16 Previous studies have shown that PEDF is highly expressed in idiopathic pulmonary fibrosis and is inducible by TGF-β in cultured human lung fibroblasts, suggesting a potential link with the pathogenesis of tissue fibrosis.17 Here, we set to determine PEDF expression in SSc skin biopsies and in TBRIIΔk-fib mice in vivo and to determine the role of PEDF expression in the antiangiogenic function of fibroblasts in vitro.

ducible by TGF-β in cultured human lung fibroblasts, suggesting a potential link with the pathogenesis of tissue fibrosis.17 Here, we set to determine PEDF expression in SSc skin biopsies and in TBRIIΔk-fib mice in vivo and to determine the role of PEDF expression in the antiangiogenic function of fibroblasts in vitro. Materials and methods Patient samples and patient skin biopsies Skin biopsies from nine patients with early diffuse cutaneous systemic sclerosis (dcSSc)18 and nine healthy controls were obtained at the SSc clinic within the Leeds Institute of Rheumatic and Musculoskeletal Medicine (UK) and the Rheumatology Unit in L’Aquila (Italy). Biopsies were taken with full informed consent as approved by National Research Ethics Service (NRES) Committee (REC 10/H1306/88) and the local ethical committee in University of L’Aquila, and processed as described in detail in online supplementary methods. 10.1136/annrheumdis-2017-212120.supp1Supplementary file 1

Materials and methods Patient samples and patient skin biopsies Skin biopsies from nine patients with early diffuse cutaneous systemic sclerosis (dcSSc)18 and nine healthy controls were obtained at the SSc clinic within the Leeds Institute of Rheumatic and Musculoskeletal Medicine (UK) and the Rheumatology Unit in L’Aquila (Italy). Biopsies were taken with full informed consent as approved by National Research Ethics Service (NRES) Committee (REC 10/H1306/88) and the local ethical committee in University of L’Aquila, and processed as described in detail in online supplementary methods. 10.1136/annrheumdis-2017-212120.supp1Supplementary file 1 Cell culture Dermal fibroblasts (FBs) from patients with dcSSc and controls were isolated as previously described19 and detailed in online supplementary methods. Human dermal microvascular endothelial cells (MVECs) were purchased from PromoCell, UK, cultured in Endothelial Cell Growth Medium BulletKit (Lonza, Slough, UK) and used at passages 2–4. Human umbilical vein endothelial cells (HUVECs) were obtained from TCS Cellworks, cultured in fully supplemented human large vessel endothelial cell medium (TCS Cellworks, UK) and used at passages 3–4. Human epidermal melanocytes (HEMs) were purchased from ScienceCell, cultured in melanocyte medium supplemented with melanocyte growth supplement (TCS Cellworks) and used at passages 3–4. All cells were kept at 37°C in a humidified atmosphere of 5% CO2.

dothelial cell medium (TCS Cellworks, UK) and used at passages 3–4. Human epidermal melanocytes (HEMs) were purchased from ScienceCell, cultured in melanocyte medium supplemented with melanocyte growth supplement (TCS Cellworks) and used at passages 3–4. All cells were kept at 37°C in a humidified atmosphere of 5% CO2. hTERT immortalisation and transduction with lentiviral short-hairpin RNA For immortalisation of primary FBs, pBabe human telomerase reverse transcriptase (hTERT) puromycin retrovirus was employed following standard protocols (described in detail in online supplementary methods). Cav-1 and PEDF expression were silenced by transduction with shRNAmir GIPZ lentiviruses (Open Biosystems, Surrey, UK) following manufacturer’s instructions (online supplementary methods). Culture treatments To evaluate the effect of TGF-β on PEDF and COLA1, cells were grown to confluence in six-well culture plates in dulbecco modified eagle medium (DMEM) 10% fetal calf serum (FCS), serum starved in DMEM 0.5% FCS for 24 hours and stimulated in the presence of 10 ng/mL recombinant human (rh) TGF-β1 (Sigma, USA) for 48 hours. Ascorbic acid (40 µg/mL) was used to optimise collagen production.20

confluence in six-well culture plates in dulbecco modified eagle medium (DMEM) 10% fetal calf serum (FCS), serum starved in DMEM 0.5% FCS for 24 hours and stimulated in the presence of 10 ng/mL recombinant human (rh) TGF-β1 (Sigma, USA) for 48 hours. Ascorbic acid (40 µg/mL) was used to optimise collagen production.20 RNA isolation and RT-PCR analysis Total RNA was isolated using the RNeasy Mini Kit (Qiagen, USA) according to the manufacturer’s instructions. One microgram of total RNA from each sample was retrotranscribed to first-strand cDNA using the SuperScript III One Step RT-PCR system (Invitrogen, UK). Quantitative RT-PCR was performed in triplicates using SYBR Green RT-PCR Master Mix Kit and the ABI PRISM 7500 Fast Real Time PCR System (Applied Biosystems) with the following primers: PEDF 5′-TGTCTCCAACTTCGGCTATG-3′  (forward) and 5′-AGTAGAGAGCCCGGTGAATG-3′ (reverse), Cav-1 5′-CGACCCTAAACACCTCAACGA-3′ (forward) and 5′-TCCCTTCTGGTTCTGTCA-3′ (reverse). Quantification was performed using the comparative CT (cycle-threshold) method employing ribosomal 18S as a housekeeping gene. Protein quantification and immunodepletion Secreted PEDF and Collagen I were detected by immunoblotting in cultured cells and supernatants and by ELISA in patient sera as described in online supplementary methods. Immunodepletion for PEDF was conducted using monoclonal mouse anti-PEDF (Chemicon-Millipore, Clone 10F12.2). Full experimental procedures are described in detail in online supplementary methods.

immunoblotting in cultured cells and supernatants and by ELISA in patient sera as described in online supplementary methods. Immunodepletion for PEDF was conducted using monoclonal mouse anti-PEDF (Chemicon-Millipore, Clone 10F12.2). Full experimental procedures are described in detail in online supplementary methods. Angiogenesis assays The organotypic co-culture assay21–23 was performed with MVECs or HUVEC and either primary or hTERT immortalised fibroblasts stably infected with lentivirus coding for either non-silencing control, Cav-1 or PEDF short-hairpin (sh) RNAs as described in detail in online supplementary methods. Number of tubules and total tubule length were analysed using the Angiosys software (TCS Cellworks). Reduced growth factor matrigel (VWR, UK) was used to perform angiogenesis matrigel assays in HUVEC employing supernatants from cultured FBs (online supplementary methods). FACS analysis and proliferation assay For PEDF detection using FACS analysis, fibroblasts were incubated in the presence of a protein transport inhibitor GolgiPlug (BD Biosciences) for 12 hours according to manufacturer’s instructions. The cells were then stained on ice with rabbit polyclonal anti-PEDF-PECy5.5 (Bioss) following fixation and permeabilisation, and analysed against the corresponding isotype control using BD FACSDiva software V.6.0 BDTM LSR II flow cytometer.

lgiPlug (BD Biosciences) for 12 hours according to manufacturer’s instructions. The cells were then stained on ice with rabbit polyclonal anti-PEDF-PECy5.5 (Bioss) following fixation and permeabilisation, and analysed against the corresponding isotype control using BD FACSDiva software V.6.0 BDTM LSR II flow cytometer. For determination of HUVEC proliferation, cells were labelled with the carbocyfluorescein succinimidyl ester dye analogue, CellTrace Violet (Invitrogen). Prior to co-culture experiments, in order to track cell division following co-culture and for accurate gating, CD90-PEvio770 and CD31-APC (Miltenyi biotec) were used to exclude potential contamination of HUVEC with co-cultured fibroblasts; 7-aminoactinomycin D was used as a viability marker. Cell division frequency and proliferation indices from list mode data were determined using proliferation wizard of ModFit software V.3.2 (Verity Software House, Topsham, ME, USA).

c) were used to exclude potential contamination of HUVEC with co-cultured fibroblasts; 7-aminoactinomycin D was used as a viability marker. Cell division frequency and proliferation indices from list mode data were determined using proliferation wizard of ModFit software V.3.2 (Verity Software House, Topsham, ME, USA). Immunohistochemistry Immunohistochemistry (IHC) analysis of human skin biopsies was performed on 3 µm paraffin sections using mouse monoclonal anti-PEDF (Clone 10F12.2; Millipore, UK), rabbit polyclonal anti-CD31, rabbit polyclonal anti-α-smooth muscle actin (SMA) (Abcam, UK) and a rabbit polyclonal anti-Cav-1 antibody (Santa Cruz, UK). Detailed procedure of the two-step staining is described in the online supplementary methods. The number of positive cells was counted by two pathologists, blinded to tissue source and expressed as the mean of two observations for each sample. For mouse skin biopsies, we employed rabbit polyclonal anti-PEDF (Aviva Systems Biology; Insight Biotech, UK), rabbit polyclonal anti-CD31 (Santa Cruz) and rabbit polyclonal anti-Cav-1 (Santa Cruz) antibodies. All sections were imaged using an Axioplan Zeiss light microscope equipped with an AxioCam digital camera.

le. For mouse skin biopsies, we employed rabbit polyclonal anti-PEDF (Aviva Systems Biology; Insight Biotech, UK), rabbit polyclonal anti-CD31 (Santa Cruz) and rabbit polyclonal anti-Cav-1 (Santa Cruz) antibodies. All sections were imaged using an Axioplan Zeiss light microscope equipped with an AxioCam digital camera. TβRIIΔk-fib animal model TβRIIΔk-fib transgenic mice were provided by Professor C. Denton of UCL Medical School Centre for Rheumatology and Connective Tissue Diseases, London, UK. The generation of TβRIIΔk-fib transgenic (TG) mice has been described previously.7 Mice were genotyped by PCR using LacZ primers and an internal control. Experiments were performed on three transgenic mice aged 6 weeks and compared with sex-matched littermate controls. Animals were housed in a conventional clean facility, with access to food and water ad libitum. Strict adherence to institutional guidelines was practised, under full local ethics committee and Home Office approval.

were performed on three transgenic mice aged 6 weeks and compared with sex-matched littermate controls. Animals were housed in a conventional clean facility, with access to food and water ad libitum. Strict adherence to institutional guidelines was practised, under full local ethics committee and Home Office approval. Results PEDF expression is increased in SSc patient biopsies compared with healthy donors In healthy skin, PEDF is expressed mainly in the germinal layer of the epidermis (figure 1A). In contrast, in early diffuse SSc, we observed a strong staining for PEDF within the lower reticular dermis, both in fibroblasts and around blood vessels (figure 1A). Double IHC studies revealed that strong positivity of PEDF correlated with strong positivity of α-SMA (figure 1A and online supplementary figure 1A, B), indicating increased number of myofibroblasts positive for PEDF in SSc biopsies. Interestingly, we also observed positivity for PEDF in vascular endothelial cells identified by CD31 staining as well as in perivascular cells (figure 1A). Quantification of PEDF-positive dermal fibroblasts within the reticular connective tissue showed significantly higher PEDF positivity in SSc fibroblasts compared with healthy control fibroblasts (figure 1B). This was accompanied by the already described decreased abundance of blood vessels in the SSc biopsies compared with healthy controls (figure 1C). Consistent with these findings, SSc skin biopsies showed 5.5-fold increase in PEDF mRNA expression as assessed by qRT-PCR (P<0.01, n=5, online supplementary figure 1D). Following this observation, we have measured PEDF concentration in 38 patients with dcSSc and 34 healthy controls from our observational cohort but found no statistically significant difference in serum levels of PEDF (P=0.87).

n PEDF mRNA expression as assessed by qRT-PCR (P<0.01, n=5, online supplementary figure 1D). Following this observation, we have measured PEDF concentration in 38 patients with dcSSc and 34 healthy controls from our observational cohort but found no statistically significant difference in serum levels of PEDF (P=0.87). 10.1136/annrheumdis-2017-212120.supp2Supplementary file 2

n PEDF mRNA expression as assessed by qRT-PCR (P<0.01, n=5, online supplementary figure 1D). Following this observation, we have measured PEDF concentration in 38 patients with dcSSc and 34 healthy controls from our observational cohort but found no statistically significant difference in serum levels of PEDF (P=0.87). 10.1136/annrheumdis-2017-212120.supp2Supplementary file 2 Figure 1 Pigment epithelium-derived factor (PEDF) expression is increased in systemic sclerosis (SSc) skin in vivo and it is inducible by transforming growth factor beta (TGF-β) in vitro. (A) Images depict representative forearm skin biopsies from healthy controls (HC) and patients with SSc (SSc) double stained for PEDF and α-smooth muscle actin (SMA), or PEDF and CD31 as indicated. Arrowheads point to spindle-shaped PEDF and SMA-positive cells within the dermis. For a larger field of stained skin section, see online supplementary figure 1A, B. Scale bars, 100 µm. (B) Dot blots show counts of PEDF-positive fibroblasts in HC and SSc samples; bars represent mean values±SEM (n=25 microscopic fields per biopsy from five different biopsies). (C) Dot blots show blood vessel counts in HC and SSc samples; bars represent average values±SEM (n=25 microscopic fields per biopsy from five different biopsies). Data on additional patient samples are shown in online supplementary table 1. (D) Histogram shows PEDF mRNA levels by RT-PCR in isolated HC fibroblasts (HC-FBs) and SSc fibroblasts (SSc-FBs); bars represent mean values±SEM (n=9 samples from each HC and SSc). Where indicated, cells were treated with 10 ng/mL TGF-β for 24 hours; HEM, human epithelial melanocytes positive control. *P<0.05, **P<0.01 by unpaired t-test. (E) Western blot shows PEDF in supernatants (SUP) collected from HC-FBs and SSc-FBs cultured in the presence or absence of TGF-β (10 ng/mL for 48 hours).

SSc). Where indicated, cells were treated with 10 ng/mL TGF-β for 24 hours; HEM, human epithelial melanocytes positive control. *P<0.05, **P<0.01 by unpaired t-test. (E) Western blot shows PEDF in supernatants (SUP) collected from HC-FBs and SSc-FBs cultured in the presence or absence of TGF-β (10 ng/mL for 48 hours). 10.1136/annrheumdis-2017-212120.supp3Supplementary file 3

SSc). Where indicated, cells were treated with 10 ng/mL TGF-β for 24 hours; HEM, human epithelial melanocytes positive control. *P<0.05, **P<0.01 by unpaired t-test. (E) Western blot shows PEDF in supernatants (SUP) collected from HC-FBs and SSc-FBs cultured in the presence or absence of TGF-β (10 ng/mL for 48 hours). 10.1136/annrheumdis-2017-212120.supp3Supplementary file 3 PEDF expression in SSc and healthy fibroblasts is induced by TGF-β To determine whether the increase in PEDF expression resulted from upregulation in gene expression, we quantified PEDF mRNA levels by RT-PCR analysis in primary dermal FBs. FBs from patients with SSc expressed approximately 5-fold higher PEDF mRNA levels compared with HC-FBs (figure 1D). Concordantly, there was over 2-fold higher levels of intracellular PEDF protein in SSc-FBs compared with HC-FBs, as assessed by FACS mean fluorescence intensity for PEDF (online supplementary figure 1C). In order to establish whether the upregulation of PEDF expression in SSc-FBs resulted from increased TGF-β signalling, we performed RT-PCR on healthy control and patient isolated fibroblasts following 48 hours of stimulation with hrTGF-β. hrTGF-β induced PEDF expression in HC-FBs by 9-fold compared with 1.6-fold increase in SSc fibroblasts (figure 1D). The levels of PEDF mRNA expression following TGF-β stimulation were comparable in HC and SSc fibroblasts. Concordantly, mean fluorescent intensity by FACS for intracellular PEDF was also increased by 2.4-fold following TGF-β stimulation (online supplementary figure 1C). Western blot analysis of supernatants from fibroblast cultures confirmed that rhTGF-β treatment increased the levels of secreted PEDF, and this was associated with an increased secretion of collagen 1 (figure 1E) as shown previously.24

increased by 2.4-fold following TGF-β stimulation (online supplementary figure 1C). Western blot analysis of supernatants from fibroblast cultures confirmed that rhTGF-β treatment increased the levels of secreted PEDF, and this was associated with an increased secretion of collagen 1 (figure 1E) as shown previously.24 SSc fibroblasts suppress angiogenesis in a PEDF-dependent manner To determine whether the observed increased expression of PEDF by dermal fibroblasts in SSc skin biopsies could contribute to defective angiogenesis, we performed endothelial-fibroblast organotypic angiogenesis in vitro assays in which endothelial cells form tubules highly reminiscent of capillaries formed during angiogenesis in vivo, embedded in natural matrix produced by the fibroblasts.21–23 Use of rhPEDF in this assay confirmed in vitro the known antiangiogenic effect of PEDF (online supplementary figure 2). In the same model, co-culture of MVECs with SSc-FBS showed a 60% decrease in the number of tubules and 90% decrease in total tubule length compared with HC-FBs (figure 2A, B). Similar results were obtained when HUVECs were co-cultured with HC-FBs or SSc-FBs (figure 2C, D), hence HUVECs were employed in subsequent assays. Knockdown of PEDF in SSc-FBs by means of lentiviral shRNA (figure 2E), followed by co-culture with ECs, rescued the number of tubules and total tubule length by 1.9-fold and 2.5-fold, respectively, compared with SScFBs infected with control lentivirus harbouring non-silencing shRNA (figure 2F, G). 10.1136/annrheumdis-2017-212120.supp4Supplementary file 4

SSc fibroblasts suppress angiogenesis in a PEDF-dependent manner To determine whether the observed increased expression of PEDF by dermal fibroblasts in SSc skin biopsies could contribute to defective angiogenesis, we performed endothelial-fibroblast organotypic angiogenesis in vitro assays in which endothelial cells form tubules highly reminiscent of capillaries formed during angiogenesis in vivo, embedded in natural matrix produced by the fibroblasts.21–23 Use of rhPEDF in this assay confirmed in vitro the known antiangiogenic effect of PEDF (online supplementary figure 2). In the same model, co-culture of MVECs with SSc-FBS showed a 60% decrease in the number of tubules and 90% decrease in total tubule length compared with HC-FBs (figure 2A, B). Similar results were obtained when HUVECs were co-cultured with HC-FBs or SSc-FBs (figure 2C, D), hence HUVECs were employed in subsequent assays. Knockdown of PEDF in SSc-FBs by means of lentiviral shRNA (figure 2E), followed by co-culture with ECs, rescued the number of tubules and total tubule length by 1.9-fold and 2.5-fold, respectively, compared with SScFBs infected with control lentivirus harbouring non-silencing shRNA (figure 2F, G). 10.1136/annrheumdis-2017-212120.supp4Supplementary file 4 Figure 2 Suppression of angiogenesis in an organotypic co-culture assay by systemic sclerosis (SSc) fibroblasts is reversed by pigment epithelium-derived factor (PEDF) knockdown. (A, C) Images show representative microscopic fields from co-culture assays of human dermal microvascular endothelial cells (MVECs) (A) or human umbilical vein endothelial cells (HUVECs) (C) seeded onto confluent fibroblasts (FBs), healthy control (HC-FBs) or SSc (SSc-FBs), stained for the endothelial marker CD31 (fibroblasts are seen unstained in the background). Note that HUVECs reproduce the behaviour of MVECs in the organotypic assays. (B, D) Histograms show the number of tubules and total tubule length quantified using Angiosys software, represented as mean±SEM (n=12 microscopic fields at ×4 magnification from triplicate wells). (E) Representative western blot showing intracellular PEDF levels in SSc fibroblasts treated with GolgiPlug, non-silencing control (NS) or with PEDF depletion (shPEDF) by means of lentiviral short-hairpin RNA (sh). (F) Images show representative microscopic fields from co-culture assays of HUVECs seeded onto confluent SSc fibroblasts (SSc-FBs), non-silencing control (NS) or with PEDF depletion (shPEDF). (G) Quantification of the number of tubules and total tubule length represented as mean±SEM (n=12 microscopic fields at ×4 magnification from triplicate wells). **P<0.01, ***P<0.001 by unpaired t-test. Scale bars, 100 µm.

o confluent SSc fibroblasts (SSc-FBs), non-silencing control (NS) or with PEDF depletion (shPEDF). (G) Quantification of the number of tubules and total tubule length represented as mean±SEM (n=12 microscopic fields at ×4 magnification from triplicate wells). **P<0.01, ***P<0.001 by unpaired t-test. Scale bars, 100 µm. Downregulation of caveolin-1 induces PEDF expression in dermal fibroblasts TGF-β downregulates caveolin-1 expression in vitro, both at the RNA and protein level,25 26 and this has been associated with myofibroblasts’ profibrotic activation.27 28 Double IHC studies showed an abundance of caveolin-1-positive cells largely negative for PEDF in HC skin, whereas in SSc skin biopsies, the majority of fibroblasts were PEDF positive and caveolin-1 negative (figure 3A, B). Moreover, there was an overall reduction of caveolin-1 positivity in SSc skin biopsies (figure 3A–C) as previously shown.18 These results suggested an inverse correlation between PEDF and caveolin-1 expression. The increased secretion of PEDF and reduced expression of caveolin-1 was conserved in vitro. SSc fibroblasts showed increased PEDF protein levels (online supplementary figure 1) and decreased expression of caveolin-1 (figure 3D). Interestingly, the relative expression of caveolin-1 in subcultured fibroblasts (SSc 1–5) followed the same trend observed by IHC in patients 1–5 (online supplementary table 1). Stable silencing of caveolin-1 by lentiviral delivery of caveolin-1 shRNA showed 85% caveolin-1 knockdown compared with control (figure 4A). Fibroblasts with caveolin-1 knockdown displayed over 2-fold upregulation of PEDF expression compared with controls (figure 4B). Accordingly, we observed an increased secretion of PEDF in the supernatants harvested from fibroblasts with caveolin-1 knockdown compared with control (figure 4C). These results show that decreased caveolin-1 expression in dermal fibroblasts induces the expression and secretion of PEDF in vitro and that the inverse correlation of expression is observable in vivo.

on of PEDF in the supernatants harvested from fibroblasts with caveolin-1 knockdown compared with control (figure 4C). These results show that decreased caveolin-1 expression in dermal fibroblasts induces the expression and secretion of PEDF in vitro and that the inverse correlation of expression is observable in vivo. Figure 3 Decreased caveolin-1 tissue expression in systemic sclerosis (SSc) is conserved in vitro and correlates with high pigment epithelium-derived factor (PEDF) expression. (A) Images show representative forearm skin biopsies from healthy controls (HC) and patients with SSc double stained for PEDF and caveolin-1 (Cav-1). Arrowheads point to Cav-1-positive cells; note that Cav-1-positive cells show no PEDF positivity. Scale bars, 50 µm. (B) Dot plots show quantification of PEDF-positive and Cav-1-negative fibroblasts (FBs) from HC and patients with SSc. (C) Dot plots show quantification of Cav-1-positive FBs in HC and patients with SSc. Data on additional patient samples are shown in online supplementary table 1. (D) Western blots of five HC and SSc FB cultures for Cav-1 and glyceraldehyde 3-phosphate dehydrogenase (GAPDH). Dermal fibroblasts were subcultured from the same biopsies analysed by immunohistochemistry and loaded in the same order as shown in panels (B) and (C) and summarised in online supplementary table 1.

able 1. (D) Western blots of five HC and SSc FB cultures for Cav-1 and glyceraldehyde 3-phosphate dehydrogenase (GAPDH). Dermal fibroblasts were subcultured from the same biopsies analysed by immunohistochemistry and loaded in the same order as shown in panels (B) and (C) and summarised in online supplementary table 1. Figure 4 Caveolin-1 (Cav-1) knockdown stimulates pigment epithelium-derived factor (PEDF) expression in dermal fibroblasts and suppresses tubulogenesis without affecting endothelial cell proliferation. (A–C) Cav-1 knockdown stimulates PEDF expression. (A) Histogram depicts caveolin-1 knockdown (shCav-1) in dermal fibroblasts by means of lentiviral short-hairpin RNA (sh), as percentage non-silencing control (NS) by RT-PCR. Bars represent mean values±SD (n=3 independent experiments carried out in triplicates). ***P<0.001 by unpaired t-test. (B) Histogram depicts PEDF mRNA by RT-PCR with caveolin-1 knockdown (shCav-1) compared with non-silencing control (NS). Bars represent mean values±SD (n=3 independent experiments carried out in triplicates). **P<0.01 by unpaired t-test. (C) Representative western blot shows levels of PEDF in supernatants (SUP) from fibroblasts with shCav-1 compared with NS control. (D) Images show representative microscopic fields from co-culture assays of human umbilical vein endothelial cells (HUVECs) seeded onto confluent systemic sclerosis fibroblasts (SSc-FBs), non-silencing control (NS) or with PEDF depletion (shPEDF) stained by CD31. Scale bars, 100 µm. Note the decreased tubule formation with caveolin-1 knockdown in fibroblasts. (E) Histograms show number of tubules and total tubule length in (D) quantified using Angiosys software, represented as mean±SEM (n=12 microscopic fields from three different experiments). (F) Representative images from HUVEC matrigel assays with cultures treated with supernatants from dermal fibroblasts, non-silencing control (NS) or with caveolin-1 knockdown (shCav-1). Note the HUVEC monolayer organisation and reduced number of loops in shCav-1 supernatant-treated cultures compared with control. (G) Quantification of number of loops in (F) represented as mean±SD  (n=12 HPF  from three different experiments). (H) Histograms show cell generation of HUVEC preloaded with CellTrace Violet co-cultured with caveolin-1 knockdown (shCav-1) fibroblasts or non-silencing control fibroblasts (NS); PI, relative proliferation indexes. Dye dilution analysis performed using Modfit proliferation algorithm.

from three different experiments). (H) Histograms show cell generation of HUVEC preloaded with CellTrace Violet co-cultured with caveolin-1 knockdown (shCav-1) fibroblasts or non-silencing control fibroblasts (NS); PI, relative proliferation indexes. Dye dilution analysis performed using Modfit proliferation algorithm. (I) Column chart showing quantification of HUVEC generations while in co-culture with caveolin-1 knockdown (shCav-1) fibroblasts, or non-silencing control fibroblasts (NS); chart indicates some increase in proliferation at G5/G6 in caveolin-1 knockdown fibroblast co-cultures (shCav-1) compared with non-silencing control fibroblasts (NS). **P<0.01; ***P<0.001 by unpaired t-test. HPF, high power fields.

h caveolin-1 knockdown (shCav-1) fibroblasts, or non-silencing control fibroblasts (NS); chart indicates some increase in proliferation at G5/G6 in caveolin-1 knockdown fibroblast co-cultures (shCav-1) compared with non-silencing control fibroblasts (NS). **P<0.01; ***P<0.001 by unpaired t-test. HPF, high power fields. Caveolin-1 downregulation in fibroblasts inhibits angiogenesis without affecting endothelial viability and proliferation Fibroblasts with lentiviral shRNA driven caveolin-1 knockdown suppressed tubule formation in the organotypic angiogenesis assay, with a 62.5% decrease in tubule length and 58.2% decrease in total number of tubules (P<0.001 for both) (figure 4D) compared with controls. Accordingly, matrigel assays using supernatants harvested from the same cells showed nearly 50% reduction in tube formation (P<0.01) (figure 4F, G). Therefore, the inhibitory effect on tubulogenesis of caveolin-1 knockdown in fibroblasts is transferable by tissue culture supernatants, supporting the notion that it is mediated via secreted factors such as PEDF. Consistently, immunodepletion of PEDF from fibroblast supernatants, through treatment with a blocking antibody (online supplementary figure 3A), rescued tubule morphogenesis significantly in the matrigel assay (online supplementary figure 3B, C). 10.1136/annrheumdis-2017-212120.supp5Supplementary file 5

Caveolin-1 downregulation in fibroblasts inhibits angiogenesis without affecting endothelial viability and proliferation Fibroblasts with lentiviral shRNA driven caveolin-1 knockdown suppressed tubule formation in the organotypic angiogenesis assay, with a 62.5% decrease in tubule length and 58.2% decrease in total number of tubules (P<0.001 for both) (figure 4D) compared with controls. Accordingly, matrigel assays using supernatants harvested from the same cells showed nearly 50% reduction in tube formation (P<0.01) (figure 4F, G). Therefore, the inhibitory effect on tubulogenesis of caveolin-1 knockdown in fibroblasts is transferable by tissue culture supernatants, supporting the notion that it is mediated via secreted factors such as PEDF. Consistently, immunodepletion of PEDF from fibroblast supernatants, through treatment with a blocking antibody (online supplementary figure 3A), rescued tubule morphogenesis significantly in the matrigel assay (online supplementary figure 3B, C). 10.1136/annrheumdis-2017-212120.supp5Supplementary file 5 Viability of endothelial cells (CD31pos/CD90neg) in the organotypic angiogenesis assay was comparable between the two experimental conditions (online supplementary figure 3D). Similarly, the proliferation index by the dye dilution method of endothelial cells was comparable in the two conditions (figure 4H, I). These data indicate that the impairment in angiogenesis, mediated by fibroblasts with caveolin-1 knockdown, does not result from decreased endothelial cell viability or proliferation.

Similarly, the proliferation index by the dye dilution method of endothelial cells was comparable in the two conditions (figure 4H, I). These data indicate that the impairment in angiogenesis, mediated by fibroblasts with caveolin-1 knockdown, does not result from decreased endothelial cell viability or proliferation. TGF-β signalling suppresses caveolin-1 expression and stimulates PEDF expression in vivo To investigate whether the findings from the patient sample analysis and culture systems hold in an in vivo model of overactivation of TGF-β signalling, we employed mice with controlled overexpression of TGF-β receptors in fibroblasts driven by the col1a1 promoter (TβRIIΔk-fib). Analysis of sections from mouse skin biopsies showed reduced caveolin-1 expression in TβRIIΔk-fib transgenic mice compared with wild-type littermate controls (figure 5A). Conversely, expression of PEDF was increased in the skin of the transgenic mice (figure 5B). Importantly, increased PEDF and reduced caveolin-1 expression in TβRIIΔk-fib transgenic mice was associated with a reduction in the number of capillaries as assessed by IHC for CD31 (figure 5C).

littermate controls (figure 5A). Conversely, expression of PEDF was increased in the skin of the transgenic mice (figure 5B). Importantly, increased PEDF and reduced caveolin-1 expression in TβRIIΔk-fib transgenic mice was associated with a reduction in the number of capillaries as assessed by IHC for CD31 (figure 5C). Figure 5 Skin immunohistochemistry of transgenic mice (TβRIIΔk-fib) and wild-type (WT) mice. Images show representative mouse skin biopsies stained for caveolin-1 (Cav-1) (A), pigment epithelium-derived factor (PEDF) (B) or CD31 (C) from transgenic TβRIIΔk-fib mice and wild-type littermate controls. Note the decreased expression of caveolin-1 and increased expression of PEDF in TβRIIΔk-fib biopsies. Scale bar, 50 µm, original magnification ×20. Plots show quantification of caveolin-1-positive (A) and PEDF-positive (B) fibroblasts (FBs), and blood vessel counts (C). Bars represent average values±SD (n=9 biopsies from three mice per genotype). *P<0.05; **P<0.01; ***P<0.001 by unpaired t-test.

n TβRIIΔk-fib biopsies. Scale bar, 50 µm, original magnification ×20. Plots show quantification of caveolin-1-positive (A) and PEDF-positive (B) fibroblasts (FBs), and blood vessel counts (C). Bars represent average values±SD (n=9 biopsies from three mice per genotype). *P<0.05; **P<0.01; ***P<0.001 by unpaired t-test. Discussion Here, we demonstrate for the first time that dermal fibroblasts in SSc play a direct role in the impairment of angiogenesis via secretion of PEDF. The increased PEDF expression in dcSSc skin biopsies validates the proteomics data of dcSSc fibroblast secretome,8 and it is consistent with the increased PEDF expression observed in IPF by Cosgrove et al.17 The PEDF-positive cells were both tissue fibroblasts (for their classic spindle-shape morphology) and myofibroblasts (SMA positive). In addition, we observed high PEDF expression in cells with perivascular localisation, suggesting that in vivo, multiple cell types may contribute to aberrant secretion of PEDF in SSc. Nevertheless, we did not find an increased PEDF serum concentration in patients with dcSSc versus healthy controls, consistent with the known paracrine mode of action of PEDF. Interestingly, PEDF was found significantly more abundant in the lower dermis of patients with dcSSc, suggesting that reticular fibroblasts may play an important role in the initiation and progression of impaired angiogenesis.

Discussion Here, we demonstrate for the first time that dermal fibroblasts in SSc play a direct role in the impairment of angiogenesis via secretion of PEDF. The increased PEDF expression in dcSSc skin biopsies validates the proteomics data of dcSSc fibroblast secretome,8 and it is consistent with the increased PEDF expression observed in IPF by Cosgrove et al.17 The PEDF-positive cells were both tissue fibroblasts (for their classic spindle-shape morphology) and myofibroblasts (SMA positive). In addition, we observed high PEDF expression in cells with perivascular localisation, suggesting that in vivo, multiple cell types may contribute to aberrant secretion of PEDF in SSc. Nevertheless, we did not find an increased PEDF serum concentration in patients with dcSSc versus healthy controls, consistent with the known paracrine mode of action of PEDF. Interestingly, PEDF was found significantly more abundant in the lower dermis of patients with dcSSc, suggesting that reticular fibroblasts may play an important role in the initiation and progression of impaired angiogenesis. Here, we show for the first time that TGF-β signalling in fibroblasts suppresses angiogenesis through secretion of PEDF, and that this pathway remains active in fibroblasts explanted from SSc skin. This is consistent with a wealth of data in the literature indicating that SSc fibroblasts maintain in vitro hallmarks of TGF-β signalling activation with passaging, including increased collagen production, phosphorylation of SMAD, Jnk and ERK as well as increased expression of α-SMA.

broblasts explanted from SSc skin. This is consistent with a wealth of data in the literature indicating that SSc fibroblasts maintain in vitro hallmarks of TGF-β signalling activation with passaging, including increased collagen production, phosphorylation of SMAD, Jnk and ERK as well as increased expression of α-SMA. Another big set of evidence indicates that PEDF mediates its antiangiogenic effects through multiple mechanisms including suppression of migration via p38 signalling, induction of apoptosis through MEK5/Erk5 signalling to peroxisome proliferator-activated receptor gamma and NF-κB,9 29 30 and antagonism of vascular endothelial growth factor (VEGF) signalling via γ-secretase cleavage of VEGF receptors.13 31 Therefore, our data link two widely demonstrated molecular pathways and suggest an explanation of TGF-β-induced vasculopathy in SSc. Intriguingly, PEDF has been found to have an antifibrotic effect in a chemically induced model of liver fibrosis.32 33 Although we cannot exclude that the TGF-β-induced expression of PEDF observed both in SSc8 34 and idiopathic pulmonary fibrosis17 could represent an attempt to negatively feedback the fibrotic process, here we show that PEDF expression can ultimately contribute to the defective angiogenesis in a paracrine manner during SSc. Previously, we and others reported that SSc skin and lung biopsies show a decreased expression of caveolin-1 when compared with healthy controls, which is associated with tissue fibrosis.18 25–28

Intriguingly, PEDF has been found to have an antifibrotic effect in a chemically induced model of liver fibrosis.32 33 Although we cannot exclude that the TGF-β-induced expression of PEDF observed both in SSc8 34 and idiopathic pulmonary fibrosis17 could represent an attempt to negatively feedback the fibrotic process, here we show that PEDF expression can ultimately contribute to the defective angiogenesis in a paracrine manner during SSc. Previously, we and others reported that SSc skin and lung biopsies show a decreased expression of caveolin-1 when compared with healthy controls, which is associated with tissue fibrosis.18 25–28 Caveolin-1 plays a bidirectional role in endothelial cells during angiogenesis (reviewed in Sowa35) by promoting36 37 or inhibiting38 blood vessel formation; however, the indirect effects on angiogenesis of low caveolin-1 levels in tissue fibroblasts were never studied before. The current study shows that similar to previous data on α-SMA and collagen, also PEDF increased expression in vivo is conserved in subcultured dermal fibroblasts consistent with the known sustained TGF-β signalling of SSc fibroblasts in vitro. Most importantly, we show that the inverse correlation in the expression of caveolin-1 and PEDF is associated with the decreased number of capillaries observed in SSc. Further, we show a causative link between caveolin-1 decreased expression and PEDF secretion, which in turn suppresses angiogenesis in vitro, without affecting cell viability or proliferation. Altogether, the rescue experiments we have performed using PEDF knockdown in SSc fibroblasts and PEDF blocking antibody in SSc supernatants (figure 2F, G and online supplementary figure 3) clearly show the importance of PEDF in the antiangiogenic phenotype of these cells.

hout affecting cell viability or proliferation. Altogether, the rescue experiments we have performed using PEDF knockdown in SSc fibroblasts and PEDF blocking antibody in SSc supernatants (figure 2F, G and online supplementary figure 3) clearly show the importance of PEDF in the antiangiogenic phenotype of these cells. Further, here we show that the inverse relationship between caveolin-1 and PEDF expression observed in the patient samples is recapitulated in the TβRIIΔk-fib transgenic mice with overactivation of TGF-β signalling in tissue fibroblasts.7 Consistent with this observation and the antiangiogenic function of PEDF, we also noted a reduced capillary density in the TβRIIΔk-fib mouse skin biopsies. Functional experiments will be necessary to determine whether this transgenic line can be used as preclinical model of TGF-β-induced vasculopathy and targeting of this signalling axis in SSc.

ation and the antiangiogenic function of PEDF, we also noted a reduced capillary density in the TβRIIΔk-fib mouse skin biopsies. Functional experiments will be necessary to determine whether this transgenic line can be used as preclinical model of TGF-β-induced vasculopathy and targeting of this signalling axis in SSc. Overall, the findings in this study strongly support the notion that TGF-β is involved in the pathogenesis of vasculopathy in SSc, and establish a causative link between caveolin-1 downregulation in tissue fibroblasts, PEDF expression and defective angiogenesis (figure 6). While this study focuses on dcSSc, in follow-up studies it would be interesting to investigate the relationship between PEDF, caveolin-1 and angiogenesis in limited cutaneous SSc (lcSSc). It is worth noting that lcSSc skin biopsies do not show the hallmarks of TGF-β activation seen in dcSSc,39 and therefore it would be very interesting to investigate whether the vasculopathy observed in lcSSc is driven by a non-TGF-β-related pathway.

en PEDF, caveolin-1 and angiogenesis in limited cutaneous SSc (lcSSc). It is worth noting that lcSSc skin biopsies do not show the hallmarks of TGF-β activation seen in dcSSc,39 and therefore it would be very interesting to investigate whether the vasculopathy observed in lcSSc is driven by a non-TGF-β-related pathway. Figure 6 Mechanistic model depicting the relationship between caveolin-1 (Cav-1) and pigment epithelium-derived factor (PEDF). Transforming growth factor beta (TGF-β) signalling strength is autoregulated by Cav-1-dependent TGF-β receptor internalisation.40 Caveolin-1 downregulation and potentiation of TGF-β signalling promotes PEDF transcription, expression and secretion by dermal fibroblasts, suppressing angiogenesis in systemic sclerosis (SSc). Exposure to high levels of TGF-β and receptor overactivation in SSc sustains caveolin-1 downregulation at the transcriptional level,27 thus promoting further PEDF expression and impairment of angiogenesis. VL and JE contributed equally. Handling editor: Tore K Kvien Contributors: VL, JEC, YME-S, MS, GG, ECD-S and FE contributed to the experimental data collection and analysis. GA and PC contributed to the identification and collection of patient clinical data and biosamples. PE, CPD, RG and GM contributed to data revision, manuscript revision and discussion. VL and JEC drafted the manuscript. FDG ideated the experimental plan and supervised all aspects of research. FDG and GM critically reviewed data and their analysis. Funding: This study was partially funded by NIHR CDF to FDG and EULAR ODP grant to FDG.

Contributors: VL, JEC, YME-S, MS, GG, ECD-S and FE contributed to the experimental data collection and analysis. GA and PC contributed to the identification and collection of patient clinical data and biosamples. PE, CPD, RG and GM contributed to data revision, manuscript revision and discussion. VL and JEC drafted the manuscript. FDG ideated the experimental plan and supervised all aspects of research. FDG and GM critically reviewed data and their analysis. Funding: This study was partially funded by NIHR CDF to FDG and EULAR ODP grant to FDG. Competing interests: None declared. Ethics approval: The study was approved by the NHS REC 10/H1306/88 and Institutional Review Board of University of L’Aquila. Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: There are no additional or unpublished data that need to be shared as part of this study.

ralia and Pfizer, and speaker fees from Actelion. AR receives funding from AstraZeneca. CPD has done consultancy for GSK, Actelion, Bayer, Inventiva and Merck-Serono, received research grant funding from GSK, Actelion, CSL Behring and Inventiva, received speaker’s fees from Bayer and given trial advice to Merck-Serono. Ethics approval: The ethics committee of each participating centre approved the study. Provenance and peer review: Not commissioned; externally peer reviewed.

Introduction Patients with the diffuse cutaneous subtype of systemic sclerosis (dcSSc) have high morbidity and mortality, associated with the degree of severity of skin fibrosis/thickening as assessed by the modified Rodnan skin score (mRSS).1 2 The mRSS, as well as being a key clinical tool that clinicians use to monitor patients in everyday clinical practice, is usually the primary end point in randomised controlled trials (RCTs) of dcSSc. These trials pose particular challenges first because dcSSc is a rare disease, and second because mRSS tends to rapidly progress over time (usually within the first 3–5 years), but then to ‘plateau’ and often subsequently fall,3 probably contributing to why several treatments associated with benefit in open-label or observational studies have not conferred benefit in RCTs.4–7 Ideally, we need to be able to predict which patients are likely to progress in terms of mRSS and recruit from this subset into RCTs. Most RCTs have restricted inclusion to patients with early disease (some within 18 months of onset of skin thickening,5 8 others within 3–5 years9–12). More recently, it has been suggested that an upper mRSS cut-off could further enrich the cohort for worsening skin,13 14 with 22 as a proposed level.13 However, the stricter the inclusion criteria, inevitably the more difficult it will be to recruit. This is a key issue: recent advances are driving new approaches to therapy, and recruitment is now increasingly difficult with competing studies.

r enrich the cohort for worsening skin,13 14 with 22 as a proposed level.13 However, the stricter the inclusion criteria, inevitably the more difficult it will be to recruit. This is a key issue: recent advances are driving new approaches to therapy, and recruitment is now increasingly difficult with competing studies. The European Scleroderma Observational Study (ESOS)15 was a prospective observational study of treatment outcome in 326 patients with early dcSSc. Patients were assessed every 3 months for 12–24 months (most for 24 months), with mRSS documented at each visit. Thus, ESOS provided a unique opportunity to perform a detailed study of mRSS trajectory over time in a large multinational cohort with very early disease (median disease duration from onset of skin thickening: 11.9 months). Our aim was twofold: (1) for the practising clinician, to identify and describe (in the ESOS cohort) patients with progressive skin thickness; and (2) for the clinical trialist, to derive prediction models for progression over 12 months, in order to inform/maximise recruitment into future RCTs.

hickening: 11.9 months). Our aim was twofold: (1) for the practising clinician, to identify and describe (in the ESOS cohort) patients with progressive skin thickness; and (2) for the clinical trialist, to derive prediction models for progression over 12 months, in order to inform/maximise recruitment into future RCTs. Methods ESOS study design and patients This is described fully elsewhere15: patients with early dcSSc were recruited into a prospective, observational cohort study comparing the effectiveness of four different treatment protocols. The main inclusion criteria were early dcSSc (skin involvement extending proximal to elbow or knee and/or involving trunk,16 and within 3 years of the onset of skin thickening as judged by physician at screening visit) and age >18 years. Patients attended every 3 months for 12–24 months. The primary outcome measure was the mRSS. Demographic and clinical characteristics including age, gender, smoking habit, ethnicity, antibody status (antitopoisomerase-1 (anti-Scl-70, ‘TOPO’), anti-RNA polymerase III (‘Pol3’), anticentromere (‘ACA’)) and presence of visceral organ involvement were recorded for all patients.15 There were 326 patients from 50 centres (19 countries) who were recruited: 65 started on methotrexate, 118 on mycophenolate mofetil, 87 on cyclophosphamide and 56 no immunosuppressant. Four patients who were found postrecruitment to have a baseline duration of skin thickening >36 months (up to 44.6) were retained (a subsidiary analysis verified the robustness of our predictive models to their inclusion). Because progression status did not significantly differ between treatment groups, mRSS trajectories were analysed irrespective of treatment protocol (online supplementary table S1). Each patient gave written informed consent.

d (a subsidiary analysis verified the robustness of our predictive models to their inclusion). Because progression status did not significantly differ between treatment groups, mRSS trajectories were analysed irrespective of treatment protocol (online supplementary table S1). Each patient gave written informed consent. 10.1136/annrheumdis-2017-211912.supp1Supplementary file 1 Definition of progressive patients Disease progression was defined in terms of mRSS worsening, in line with most recent RCTs. For the univariate analysis and predictive models, patients with progressive disease (‘progressors’) are defined as those with a 5-unit and 25% increase in their mRSS between baseline and their highest subsequent score. This threshold is generally considered to reflect meaningful change in mRSS progression,17 thus enabling model comparisons.13 14 18 We considered only peaks occurring during the first 12±3 months after baseline, using all 3-monthly observations. The time window was chosen because it is considered an appropriate period to detect clinically meaningful changes in the skin score.19 Most cases of progression occurred early: extending the time period to 24 months would have added only four additional ‘progressors’ and would have lost comparability with other published models of progression which examined a 12-month window.13 14 18 To distinguish between non-progressors and patients with insufficient data to describe their status, data requirements were set up as detailed in table 1 footnote (*).

Definition of progressive patients Disease progression was defined in terms of mRSS worsening, in line with most recent RCTs. For the univariate analysis and predictive models, patients with progressive disease (‘progressors’) are defined as those with a 5-unit and 25% increase in their mRSS between baseline and their highest subsequent score. This threshold is generally considered to reflect meaningful change in mRSS progression,17 thus enabling model comparisons.13 14 18 We considered only peaks occurring during the first 12±3 months after baseline, using all 3-monthly observations. The time window was chosen because it is considered an appropriate period to detect clinically meaningful changes in the skin score.19 Most cases of progression occurred early: extending the time period to 24 months would have added only four additional ‘progressors’ and would have lost comparability with other published models of progression which examined a 12-month window.13 14 18 To distinguish between non-progressors and patients with insufficient data to describe their status, data requirements were set up as detailed in table 1 footnote (*). Table 1 Characteristics of progressors and non-progressors according to clinical features and autoantibody status (at baseline)

Definition of progressive patients Disease progression was defined in terms of mRSS worsening, in line with most recent RCTs. For the univariate analysis and predictive models, patients with progressive disease (‘progressors’) are defined as those with a 5-unit and 25% increase in their mRSS between baseline and their highest subsequent score. This threshold is generally considered to reflect meaningful change in mRSS progression,17 thus enabling model comparisons.13 14 18 We considered only peaks occurring during the first 12±3 months after baseline, using all 3-monthly observations. The time window was chosen because it is considered an appropriate period to detect clinically meaningful changes in the skin score.19 Most cases of progression occurred early: extending the time period to 24 months would have added only four additional ‘progressors’ and would have lost comparability with other published models of progression which examined a 12-month window.13 14 18 To distinguish between non-progressors and patients with insufficient data to describe their status, data requirements were set up as detailed in table 1 footnote (*). Table 1 Characteristics of progressors and non-progressors according to clinical features and autoantibody status (at baseline) Characteristics Progressor, n=66 (22.5%) Non-progressor, n=227 (77.5%) P Total, n=293 (100%) Missing at baseline, n (%) mRSS (0–51) 19 (16–23) 21 (16–27) 0.030 21 (16–26) 0 (0) Months since onset of skin thickening 8.1 (4.7–16.0) 12.6 (8.1–22.0) 0.001 12.0 (7.0–21.0) 14 (4.8) Pulmonary fibrosis, n (%) 9 (13.6) 31 (13.7) 1 40 (13.7) 0 (0) FVC (% predicted) 87.5 (72.0–101.0) 91.0 (75.0–102.0) 0.129 90.0 (75.0–102.0) 16(5.5) DLCO (% predicted) 62.8 (49.0–76.5) 66.0 (52.0–79.0) 0.455 65.0 (50.0–79.0) 31 (10.6) Pulmonary hypertension, n (%) 1 (1.5) 19 (8.4) 0.054† 20 (6.8) 1 (0.3) Antitopoisomerase (anti-Scl70) (TOPO), n (%) 30 (46.2) 84 (38) 0.252 114 (39.9) 7 (2.4) Anti-RNA polymerase III (Pol3), n (%) 14 (25.9) 34 (18.5) 0.249 48 (20.2) 55 (18.8) Anticentromere (ACA), n (%) 5 (7.7) 14 (6.4) 0.777 19 (6.7) 8 (2.7) No autoantibodies (TOPO, Pol3 or ACA), n (%) 7 (12.7) 52 (28.4) 0.020 59 (24.8) 55 (18.8) Median (IQR) unless otherwise indicated.

6.2) 84 (38) 0.252 114 (39.9) 7 (2.4) Anti-RNA polymerase III (Pol3), n (%) 14 (25.9) 34 (18.5) 0.249 48 (20.2) 55 (18.8) Anticentromere (ACA), n (%) 5 (7.7) 14 (6.4) 0.777 19 (6.7) 8 (2.7) No autoantibodies (TOPO, Pol3 or ACA), n (%) 7 (12.7) 52 (28.4) 0.020 59 (24.8) 55 (18.8) Median (IQR) unless otherwise indicated. P values refer to the Kruskal-Wallis (for continuous variables) or Fisher’s test (for categorical variables). This table compares the distribution of patient characteristics at baseline between progressors and non-progressors, using the subset of 293 for whom the progression status is known. To distinguish between non-progressors and patients with insufficient data to describe their status, data requirements were set up. If progression was not detected using all data from the first >12±3 months, patients needed at least two data points to be considered non-progressors: one at baseline and another at least 5 months after baseline. Otherwise, we considered there were not enough data to ascertain their status. The 5-month limit was chosen so that all visits in the vicinity of the 6-month study mark could be counted. †The presence of pulmonary hypertension was not included as a variable in prediction models for progression. Only one patient had pulmonary hypertension and progressed. Thus, a prediction model using mRSS, duration of skin thickening, an mRSS/duration interaction and the presence of pulmonary hypertension was too restrictive: no combinations of mRSS and duration of skin thickening enabled patients with pulmonary hypertension to progress.

ient had pulmonary hypertension and progressed. Thus, a prediction model using mRSS, duration of skin thickening, an mRSS/duration interaction and the presence of pulmonary hypertension was too restrictive: no combinations of mRSS and duration of skin thickening enabled patients with pulmonary hypertension to progress. DLCO, diffusing capacity for carbon monoxide; FVC, forced vital capacity; mRSS, modified Rodnan skin score. Univariate analysis The univariate analysis compared progressors and non-progressors according to patient characteristics using the Kruskal-Wallis (for continuous variables) or Fisher’s test (for categorical variables). To characterise the progression of skin thickening according to autoantibody status, those same tests assessed differences in distribution for certain features (such as disease duration and mRSS peak) between autoantibody groups. If a patient tested positive for an autoantibody, we assumed they did not have the other two if those data were missing. Patients with more than one autoantibody were excluded from our models. Predictive models of mRSS progression Logistic regressions were fitted to predict progression using baseline characteristics. Associations with progression (including those in table 1) and the predictive performance of single predictors were assessed to select potential covariates, resulting in different models.

If a patient tested positive for an autoantibody, we assumed they did not have the other two if those data were missing. Patients with more than one autoantibody were excluded from our models. Predictive models of mRSS progression Logistic regressions were fitted to predict progression using baseline characteristics. Associations with progression (including those in table 1) and the predictive performance of single predictors were assessed to select potential covariates, resulting in different models. Those models were then compared on the basis of the area under curve (AUC), sensitivity, specificity, positive predictive value (PPV) and accuracy at each curve’s optimal point—but also according to their simplicity and interpretability. Predictive ability can be optimistic when assessed using its own model-generating data. An additional optimism-adjusted bootstrapped AUC was therefore also computed and reported in online supplementary table S2, suggesting modest corrections.20 Calibration plots for the retained models were also assessed.21 When including autoantibodies in predictive models, certain specifications produced predicted progression probabilities that were too low for certain subgroups and were thus avoided because they were considered too restrictive to apply in practice. Consequently, patients were only classed according to their Pol3 positivity rather than having indicator variables for each autoantibody (see note (1) in figure 3 and online supplementary table S2).

re too low for certain subgroups and were thus avoided because they were considered too restrictive to apply in practice. Consequently, patients were only classed according to their Pol3 positivity rather than having indicator variables for each autoantibody (see note (1) in figure 3 and online supplementary table S2). Results Univariate analysis: associates of mRSS progression and autoantibody status The characteristics of mRSS progression are summarised in figure 1, including the increase in mRSS and the peak reached. During the study, the median number of skin scores recorded for each patient was 7 over a median follow-up of 23.4 months. There were 160 patients who had an increase in mRSS (of any magnitude) during the study (149 during the first 12 (±3) months). Figure 1 Characteristics of mRSS progression. The five histograms describe modified Rodnan skin score (mRSS) progression for all patients whose skin score during the study ever increases beyond their baseline level (n=160) and for those whose progression satisfies the 5-unit and 25% increase rule during the first 12 months (±3 months) (n=66). Here, histograms summarise the distribution of changes between baseline and peak mRSS, the mRSS value at its peak, the time elapsed between the onset of skin thickening and the recorded peak, the rate of mRSS increase per month between baseline and peak, and the time elapsed between baseline and the recorded peak. The rate of mRSS progression (in units/month) was computed by specifying individual simple linear regressions of mRSS according to time, between baseline and peak.

hickening and the recorded peak, the rate of mRSS increase per month between baseline and peak, and the time elapsed between baseline and the recorded peak. The rate of mRSS progression (in units/month) was computed by specifying individual simple linear regressions of mRSS according to time, between baseline and peak. Characteristics of progressors versus non-progressors Out of 326 patients recruited at baseline, based on the retained progression criterion, 66 (22.5%) progressed and 227 (77.5%) did not (table 1). Progression status could not be assessed in 33 patients: 16 had no postbaseline skin scores and 17 did not fulfil the data requirements to ascertain progression status (see footnote (*) of table 1). Among those 33 patients with unknown status, 12 (36.4%) died during the analysis period. At the time of recruitment, progressors had shorter disease duration than those who did not progress (median 8.1 vs 12.6 months (P=0.001)). In addition, progressors tended to start with lower skin scores, median mRSS of 19 units, compared with 21 for non-progressors (P=0.030). Nevertheless, 30.3% of progressors started with mRSS >22 units and 15.2% with mRSS >25 units (online supplementary figure S1). 10.1136/annrheumdis-2017-211912.supp2Supplementary file 2

In addition, progressors tended to start with lower skin scores, median mRSS of 19 units, compared with 21 for non-progressors (P=0.030). Nevertheless, 30.3% of progressors started with mRSS >22 units and 15.2% with mRSS >25 units (online supplementary figure S1). 10.1136/annrheumdis-2017-211912.supp2Supplementary file 2 Characteristics of mRSS progression according to autoantibody status Out of the 326 patients, 124 were TOPO+, 50 were Pol3+, 20 were ACA+, 2 were TOPO+/ACA+, 68 were autoantibody-negative and 62 could not have their status determined: in 51 cases, this was because the Pol3 test was not done (unavailable in some centres) and the patient had neither TOPO nor ACA antibodies (table 2). Table 2 Characteristics of mRSS progression according to autoantibody status

Characteristics of mRSS progression according to autoantibody status Out of the 326 patients, 124 were TOPO+, 50 were Pol3+, 20 were ACA+, 2 were TOPO+/ACA+, 68 were autoantibody-negative and 62 could not have their status determined: in 51 cases, this was because the Pol3 test was not done (unavailable in some centres) and the patient had neither TOPO nor ACA antibodies (table 2). Table 2 Characteristics of mRSS progression according to autoantibody status Autoantibody make-up Anti-TOPO-isomerase (anti-Scl70) (TOPO) Anti-RNA polymerase III (Pol3) Anticentromere (ACA) None Total P Missing at baseline, n (%) (TOPO+) (Pol3– or N/A) (ACA– or N/A) (TOPO– or N/A) (Pol3+) (ACA– or N/A) (TOPO– or N/A) (Pol3– or N/A) (ACA+) (TOPO−) (Pol3−) (ACA−) n=124 (47.3%) n=50 (19.1%) n=20 (7.6%) n=68 (26.0%) n=262 (100%) mRSS at baseline (0–51) 19 (15–25.5) 24 (19-31) 20 (17–24.5) 20 (16–24) 20 (16–26) 0.003 0 (0) mRSS peak* 26 (19.5–33.5) 35 (26–40) 29 (26–35) 24.5 (17.5–29) 27 (21–34.5) 0.001 0 (0) Difference in mRSS between baseline and peak* 5 (3–10.5) 7 (3–10) 4 (4–11) 3 (1.5–7) 5 (3–10) 0.059 0 (0) Months since onset of skin thickening (at baseline) 12.6 (6.2–21.6) 11.2 (7.8–17.9) 14.9 (5.4–24.0) 12.6 (9.2–21.9) 12.6 (7.3–21.5) 0.593 10 (3.8) Months until peak since onset of skin thickening* 21.0 (12.9–31.6) 16.3 (12.9–21.4) 29.3 (15.5–35.7) 20.1 (13.2–32) 19.0 (12.9–30.0) 0.199 5 (3.9) Months until peak since baseline* 6.4 (4.0–14.4) 5.8 (2.9–12.0) 6.5 (2.9–9.2) 6.0 (3.1–11.6) 6.2 (3.2–12.1) 0.329 0 (0) Progressor (5 points and 25% according to baseline) 29 (25.9) 14 (29.2) 4 (23.5) 7 (11.9) 54 (22.9) 0.105 26 (9.9) Median (IQR) unless otherwise indicated.

0.1 (13.2–32) 19.0 (12.9–30.0) 0.199 5 (3.9) Months until peak since baseline* 6.4 (4.0–14.4) 5.8 (2.9–12.0) 6.5 (2.9–9.2) 6.0 (3.1–11.6) 6.2 (3.2–12.1) 0.329 0 (0) Progressor (5 points and 25% according to baseline) 29 (25.9) 14 (29.2) 4 (23.5) 7 (11.9) 54 (22.9) 0.105 26 (9.9) Median (IQR) unless otherwise indicated. P values refer to the Kruskal-Wallis (for continuous variables) or Fisher’s test (for categorical variables). *For these comparisons, an unrestricted definition of progression was used, meaning that all 160 patients in the cohort with mRSS progression of any magnitude were initially considered but only 128 of those could be included because of patients with missing autoantibody data. This table includes comparisons of patient characteristics at baseline between different autoantibody groups, using the subset of 262 patients for whom the autoantibody status could be assessed. mRSS, modified Rodnan skin score; N/A, not available. At baseline, Pol3+ patients had higher mRSS than patients in the other autoantibody groups (P=0.003) despite similar disease durations (P=0.593) (table 2).

This table includes comparisons of patient characteristics at baseline between different autoantibody groups, using the subset of 262 patients for whom the autoantibody status could be assessed. mRSS, modified Rodnan skin score; N/A, not available. At baseline, Pol3+ patients had higher mRSS than patients in the other autoantibody groups (P=0.003) despite similar disease durations (P=0.593) (table 2). There was a trend for Pol3+ patients to be more likely to progress than the other subgroups: 29.2% were progressors compared with 11.9% for the ‘no autoantibody’ group (P=0.105) (table 2). Pol3+ patients experienced higher increases in mRSS between baseline and peak: median increase of 7 units, compared with 3 for the ‘no autoantibody’ group (P=0.059) (table 2). Combined with their higher mRSS starting point, this results in Pol3+ patients having the highest peaks of all autoantibody groups with a median peak of 35 units (P=0.001) (table 2). In terms of the speed of progression following onset, Pol3+ patients had the lowest observed median time to peak at 16.3 months (P=0.199) (table 2).

There was a trend for Pol3+ patients to be more likely to progress than the other subgroups: 29.2% were progressors compared with 11.9% for the ‘no autoantibody’ group (P=0.105) (table 2). Pol3+ patients experienced higher increases in mRSS between baseline and peak: median increase of 7 units, compared with 3 for the ‘no autoantibody’ group (P=0.059) (table 2). Combined with their higher mRSS starting point, this results in Pol3+ patients having the highest peaks of all autoantibody groups with a median peak of 35 units (P=0.001) (table 2). In terms of the speed of progression following onset, Pol3+ patients had the lowest observed median time to peak at 16.3 months (P=0.199) (table 2). Predictive models of mRSS progression in first year of follow-up Univariate and multivariate predictive models Online supplementary table S2 and figure 2 show the values associated with the ROC curves for the multiple models tested, and online supplementary table S3 displays the details of different selected logistic models to predict progression and the regression outputs. As a single predictor for progression, mRSS performed poorly with an AUC of 0.588 (95% CI 0.515 to 0.661). Duration of skin thickening performed better on its own, with an AUC of 0.634 (95% CI 0.553 to 0.715). A model combining mRSS, disease duration and an interaction between the two improved those univariate performances, with an AUC of 0.666 (95% CI 0.597 to 0.736). In addition, that model had a high 73.4% sensitivity, alongside its 57.2% specificity, and accurately predicted 60.9% of cases.

0.634 (95% CI 0.553 to 0.715). A model combining mRSS, disease duration and an interaction between the two improved those univariate performances, with an AUC of 0.666 (95% CI 0.597 to 0.736). In addition, that model had a high 73.4% sensitivity, alongside its 57.2% specificity, and accurately predicted 60.9% of cases. Figure 2 ROC of three selected models. Three ROC curves summarise the predictive power of three different models by plotting sensitivity with respect to 100-specificity. For each model/ROC curve, there is an optimal point (the one closest to the top-left corner) that corresponds to a threshold of predicted probability of progression. For each model, patients with a predicted probability above that threshold are predicted to progress. AUC, area under curve; mRSS, modified Rodnan skin score; Pol3, anti-RNA polymerase III; PPV, positive predictive value; ROC, receiver operating characteristic. The interaction between mRSS and disease duration indicated that future progressors presented at their first visit with earlier disease and lower skin scores, and that higher skin score usually had to be compensated by lower disease duration for progression to occur (figure 3). Graphically, this could be identified by noting that, in figure 3 (model A), the points indicating progressors were mostly contained within the triangular lower half of a rectangle.

lower skin scores, and that higher skin score usually had to be compensated by lower disease duration for progression to occur (figure 3). Graphically, this could be identified by noting that, in figure 3 (model A), the points indicating progressors were mostly contained within the triangular lower half of a rectangle. Figure 3 Rules for selecting progressive patients according to two selected models. According to each model, in order to select progressive patients, they should be selected from the area under each relevant curve. These curves are superposed over a plot of the baseline mRSS of patients with respect to their duration of skin thickening, with progressors (of at least 5 units and 25%) being highlighted. Notes to the figure: (1) Analysing patients separately according to all autoantibody groups (TOPO, Pol3, ACA, ‘no autoantibodies’) was avoided because of the small number of ACA+ patients who could be included (n=16). Another possible approach was the inclusion of two indicator variables: one for Pol3+ and another for TOPO+, meaning that ACA+ and ‘no autoantibody’ patients formed the reference group, for which the resulting model proved too restrictive: only 2 out of 75 patients in the reference group were predicted to progress. Another reason for considering Pol3 patients separately was that they were suspected from preliminary analysis to be the most clinically different group, and stratifying by Pol3 status produced a higher AUC than doing so by TOPO status. (2) Each prediction model is based on a logistic regression model, where the outcome for patient i is Yi=progression and X are a selection of covariates. Using ROC curve analysis, each model has an optimal p∗ for which, if Pr^(Yi=1|X)>p∗, the patient is predicted to progress. Each frontier in the graphs above corresponds to the combination of mRSS and disease duration points, for which Pr^(Yi=1|X)=p∗ in the domains where both predictors are defined. Therefore, if a patient is in the area under the relevant curve, she/he is predicted to progress according to the model. ACA, anticentromere; AUC, area under curve; mRSS, modified Rodnan skin score; Pol3, anti-RNA polymerase III; ROC, receiver operating characteristic; TOPO, topoisomerase.

both predictors are defined. Therefore, if a patient is in the area under the relevant curve, she/he is predicted to progress according to the model. ACA, anticentromere; AUC, area under curve; mRSS, modified Rodnan skin score; Pol3, anti-RNA polymerase III; ROC, receiver operating characteristic; TOPO, topoisomerase. Adding an indicator variable for Pol3 positivity induced further gains in the model (already including mRSS, duration and their interaction), yielding an AUC of 0.711 (95% CI 0.633 to 0.790), 60.4% sensitivity, 74.2% specificity and accurately predicting 71% of cases (online supplementary table S2). By graphical assessment, model A appeared to be better calibrated than the one including only mRSS, and model B appeared to improve on model A (online supplementary figure S2). The predicted probabilities for models A and B are summarised in online supplementary figures S3 and S4. 10.1136/annrheumdis-2017-211912.supp3Supplementary file 3

By graphical assessment, model A appeared to be better calibrated than the one including only mRSS, and model B appeared to improve on model A (online supplementary figure S2). The predicted probabilities for models A and B are summarised in online supplementary figures S3 and S4. 10.1136/annrheumdis-2017-211912.supp3Supplementary file 3 Properties of predictive models and application in practice Two models described above were retained: model A and model B, which also includes Pol3+ status (figure 3, online supplementary table S2). Their ROC curves are shown in figure 2, and each curve yielded an optimal point nearest to the top-left corner, representing a threshold probability of progression. If a patient’s predicted probability was above this threshold, it was predicted that she/he would progress. Thus, for each level of disease duration at baseline, there corresponded an entry mRSS under which a patient met that threshold (summarised and plotted in table 3 and figure 3 for the two models). For instance, using the selection rule produced by model A, a patient recruited at 9 months of skin thickening would be predicted to progress if mRSS was between 0 and 23 units. However, if a patient presented at 6 months, the mRSS would be allowed to go as high as 29. Table 3 Rules for selecting progressive patients according to two selected models

Properties of predictive models and application in practice Two models described above were retained: model A and model B, which also includes Pol3+ status (figure 3, online supplementary table S2). Their ROC curves are shown in figure 2, and each curve yielded an optimal point nearest to the top-left corner, representing a threshold probability of progression. If a patient’s predicted probability was above this threshold, it was predicted that she/he would progress. Thus, for each level of disease duration at baseline, there corresponded an entry mRSS under which a patient met that threshold (summarised and plotted in table 3 and figure 3 for the two models). For instance, using the selection rule produced by model A, a patient recruited at 9 months of skin thickening would be predicted to progress if mRSS was between 0 and 23 units. However, if a patient presented at 6 months, the mRSS would be allowed to go as high as 29. Table 3 Rules for selecting progressive patients according to two selected models If duration of skin thickening is (months) Model A Model B AUC: 0.666 Sensitivity: 73.4% Specificity: 57.2% PPV: 33.8% NPV: 87.9% Accuracy: 60.9% AUC: 0.711 Sensitivity: 60.4% Specificity: 74.2% PPV: 41.0% NPV: 86.3% Accuracy: 71.0% All patients Pol3+ patients All others mRSS should be (units) or less mRSS should be (units) or less mRSS should be (units) or less 1 51 51 51 2 51 51 51 3 43 51 51 4 37 51 37 5 33 51 28 6 29 48 23 7 27 40 20 8 25 35 18 9 23 31 16 10 22 28 15 11 21 26 14 12 20 24 13 13 19 22 13 14 18 21 12 15 18 20 12 16 17 19 11 17 17 18 11 18 16 17 11 19 16 17 10 20 15 16 10 21 15 16 10 22 15 15 10 23 15 15 10 24 14 14 9 25 14 14 9 26 14 14 9 27 14 13 9 28 13 13 9 29 13 13 9 30 13 13 9 31 13 12 9 32 13 12 9 33 13 12 8 34 12 12 8 35 12 12 8 36 12 11 8 For each duration, the required mRSS level is rounded above to the nearest integer to reflect real mRSS values.

17 11 19 16 17 10 20 15 16 10 21 15 16 10 22 15 15 10 23 15 15 10 24 14 14 9 25 14 14 9 26 14 14 9 27 14 13 9 28 13 13 9 29 13 13 9 30 13 13 9 31 13 12 9 32 13 12 9 33 13 12 8 34 12 12 8 35 12 12 8 36 12 11 8 For each duration, the required mRSS level is rounded above to the nearest integer to reflect real mRSS values. AUC, area under curve; mRSS, modified Rodnan skin score; PPV, positive predictive value. If applying this selection rule (model A) to the ESOS cohort, 139 patients (49.8% of the 279 patients included in the model) would be predicted to progress, of whom 47 actually did in the year following baseline (PPV: 33.8%). Conversely, 140 were predicted not to progress, of whom 123 did not (negative predictive value (NPV): 87.9%), whereas 17 (12.1%) did. Model B is used in the same way as model A, but accounting for Pol3 status. The curves in figure 3 (summarising selection criteria) shift across the diagonal axis to reflect that Pol3+ patients have a higher propensity to progress during the first year compared with Pol3− patients. Model B had a higher accuracy than model A (71.0%–60.9%)%). Model B, which was more specific, was also more restrictive: only 78 patients were predicted to progress, of whom 32 actually did (PPV: 41.0%). Therefore this model identified a ‘high risk’ subset of patients with a proportion of progressors 1.8 times higher than the overall cohort. In model B, 153 patients were predicted not to progress, of whom 132 did not (NPV: 86.3%), whereas 21 did (13.7%).

were predicted to progress, of whom 32 actually did (PPV: 41.0%). Therefore this model identified a ‘high risk’ subset of patients with a proportion of progressors 1.8 times higher than the overall cohort. In model B, 153 patients were predicted not to progress, of whom 132 did not (NPV: 86.3%), whereas 21 did (13.7%). The predictive power of model B was particularly strong for Pol3+ patients, for whom the sensitivity was 100% and the specificity was 70.6%.

were predicted to progress, of whom 32 actually did (PPV: 41.0%). Therefore this model identified a ‘high risk’ subset of patients with a proportion of progressors 1.8 times higher than the overall cohort. In model B, 153 patients were predicted not to progress, of whom 132 did not (NPV: 86.3%), whereas 21 did (13.7%). The predictive power of model B was particularly strong for Pol3+ patients, for whom the sensitivity was 100% and the specificity was 70.6%. Discussion The major strength of this study compared with previous recent analyses of mRSS is that this was a well-defined cohort with prospective assessment of mRSS by experienced assessors. Assessments every 3 months provide detailed insight into disease trajectory (and burden) for the practising clinician. For the clinical trialist, the time frames examined were comparable to those of recent and current RCTs, which include assessments at 24 weeks (and less) as well as at 12 months.12 22 In addition, as the data set was derived from an observational study of standard current treatments for skin, we expect that our findings are generalisable to current or future clinical trials of skin therapy in dcSSc. This is especially relevant since current trials often permit standard background therapy, as used in ESOS, to which a novel agent may be added. The key finding here was the development of a predictive model for mRSS (disease) progression which had an accuracy of 60.9% (model A), achieved by recognising that the initial skin score is a poor predictor of progression on its own and that prediction is improved by simultaneously accounting for disease duration. By including autoantibodies in this analysis, the model improved and reached an accuracy of 71.0% (model B).

n accuracy of 60.9% (model A), achieved by recognising that the initial skin score is a poor predictor of progression on its own and that prediction is improved by simultaneously accounting for disease duration. By including autoantibodies in this analysis, the model improved and reached an accuracy of 71.0% (model B). When recruiting patients into clinical trials of rare diseases, any algorithm should not be too restrictive. Higher sensitivity was favoured because it was considered more appropriate to have more inclusive models at the risk of mischaracterising non-progressors as progressors. We believe that model A will be the more useful for studies aiming for cohort enrichment, while model B will help to identify patients at higher risk for mRSS progression in a clinical setting. The use of the second model to inform patient selection into RCTs could risk over-representing Pol3+ patients, for whom the criteria to predict progression are less strict, thus yielding a sample not reflecting the overall dcSSc population.

to identify patients at higher risk for mRSS progression in a clinical setting. The use of the second model to inform patient selection into RCTs could risk over-representing Pol3+ patients, for whom the criteria to predict progression are less strict, thus yielding a sample not reflecting the overall dcSSc population. Other ‘take home messages’ were that skin score progression did occur in some patients who presented with high baseline mRSS (25 or higher), although this tended to be compensated by shorter disease durations, that Pol3+ patients tended to reach their peak mRSS earlier than other patients, and that this peak was much higher than for patients with other (or no) autoantibodies. Patients without TOPO, Pol3 or ACA autoantibodies had smaller increases in mRSS and lower peak skin scores. Our 3-monthly data allowed us to capture peaks in mRSS, which would have been ‘smoothed over’ in other studies because of less frequent data. Had we only recorded baseline and 12-month data (two observations), 53% of our cases of progression would have been missed.

smaller increases in mRSS and lower peak skin scores. Our 3-monthly data allowed us to capture peaks in mRSS, which would have been ‘smoothed over’ in other studies because of less frequent data. Had we only recorded baseline and 12-month data (two observations), 53% of our cases of progression would have been missed. Taking into account peak mRSS in defining progression (as opposed to considering only baseline and 12 month data) was therefore a major difference between ESOS and the study by Maurer et al13 who also looked at prediction of extent of skin thickening in patients with systemic sclerosis in a study of 637 patients from the EULAR Scleroderma Trials and Research group  (EUSTAR) cohort and an average follow-up time between visits of 12 months (compared with 3-monthly in ESOS). Disease duration was 42 months (therefore substantially longer than in the ESOS cohort) and baseline mRSS was 17 units (compared with a mean of 22.1 units for ESOS). ESOS had 22.5% of progressors compared with EUSTAR’s 9.7%, possibly because ESOS was an earlier cohort and the 3-monthly follow-ups made any disease progression more likely to be detected. Maurer et al 13 established that lower mRSS and shorter disease duration were associated with more progressive cases, as confirmed here, although we accept that the two studies are not strictly comparable given the differing time frames of defining ‘progressors’.13

de any disease progression more likely to be detected. Maurer et al 13 established that lower mRSS and shorter disease duration were associated with more progressive cases, as confirmed here, although we accept that the two studies are not strictly comparable given the differing time frames of defining ‘progressors’.13 However, if we do apply a 22-unit mRSS cut-off point to the ESOS cohort, its size would decrease from 326 to 189, and the share of progressors (among those with known status) would only increase from 22.5% to 26.4%. In contrast, that share (PPV) rises to 33.8% with model A and 41.0% with model B. Like Maurer et al 13 we found that skin score alone was a poor predictor for progression and that other factors including disease duration should also be considered. Dobrota et al 14 also looked at patterns of mRSS changes but focused on regression rather than progression, validating that a low baseline mRSS predicts progression.

However, if we do apply a 22-unit mRSS cut-off point to the ESOS cohort, its size would decrease from 326 to 189, and the share of progressors (among those with known status) would only increase from 22.5% to 26.4%. In contrast, that share (PPV) rises to 33.8% with model A and 41.0% with model B. Like Maurer et al 13 we found that skin score alone was a poor predictor for progression and that other factors including disease duration should also be considered. Dobrota et al 14 also looked at patterns of mRSS changes but focused on regression rather than progression, validating that a low baseline mRSS predicts progression. Our study has certain limitations. It can be very difficult to gauge onset of skin thickening (in 18 (5.5%) patients we had no data on duration of skin thickening at baseline, other than that this was under 3 years). It is likely that in some patients (especially those who steadily improve after baseline), peak mRSS occurred prior to study entry. Also, unlike the EUSTAR study,13 we have not externally validated the model, and this will be an important step before using the models widely. Among the patients with unknown progression status, 36.4% died, thus potentially inducing bias (it is likely they had progressive disease) but also mirroring the attrition occurring in clinical trials. In model B, missing data in autoantibodies (19.6%) reduce the predictive power.

t step before using the models widely. Among the patients with unknown progression status, 36.4% died, thus potentially inducing bias (it is likely they had progressive disease) but also mirroring the attrition occurring in clinical trials. In model B, missing data in autoantibodies (19.6%) reduce the predictive power. In conclusion, among patients with early dcSSc, those with shorter disease duration and lower mRSS are most likely to be ‘progressors’ with a trade-off between the two factors, and patients who are Pol3+ have the highest mRSS peaks and tend to reach peak mRSS earliest, providing a valuable message for clinicians that patients with short disease duration and Pol3+ must be especially closely monitored. Two prediction models for progressive skin thickening were derived. The model incorporating Pol3 (model B) more accurately identifies high-risk patients, but risks being too restrictive for patient selection into trials and over-representing Pol3+ patients. Both models were more flexible (for a given skin score) and more accurate than a ‘22 mRSS’ cut-off model and may offer advantages for cohort enrichment in clinical trials to ensure that the most informative patients are included. 10.1136/annrheumdis-2017-211912.supp4Supplementary file 4 10.1136/annrheumdis-2017-211912.supp5Supplementary file 5 We are grateful to Dr Holly Ennis for study set-up and for project coordination during the earlier phases of the study, and also to the members of the independent oversight board: Stephen Cole, Dinesh Khanna and Frank Wollheim. Handling editor: Tore K Kvien

10.1136/annrheumdis-2017-211912.supp4Supplementary file 4 10.1136/annrheumdis-2017-211912.supp5Supplementary file 5 We are grateful to Dr Holly Ennis for study set-up and for project coordination during the earlier phases of the study, and also to the members of the independent oversight board: Stephen Cole, Dinesh Khanna and Frank Wollheim. Handling editor: Tore K Kvien Contributors: ALH, ML, RH, LM, AJS, EB, LaC, JHWD, OD, KF, WJG, RO, MCV and CPD were members of the European Scleroderma Observational Study (ESOS) Steering Committee and designed the ESOS study. SeP and ML were responsible for the statistical analysis. ALH, RH, LaC, JHWD, OD, MV, CoA, VHO, DF, MH, MM-C, AB-G, OM, PJ, ACJ, WS, PM, FCH, ChA, MEA, ED, RM, MA, MHB, LoC, NSD, HG, PL, YA, KC, SJ, AJM, NM, UM-L, GR, MB, JR, PEC, A-LF, EH, JH, MI, JSM, JMvL, SaP, SuP, AR, JS, BC, CS, TS, DJV, CG, G-ST and CPD were principal investigators at the different sites and recruited patients. XP and GD were study coordinators. ALH, SeP, ML, RH, LM, AJS and CPD wrote the draft report, and all authors reviewed the report and provided comments. Funding: ESOS was funded by a grant from the EULAR (European League Against Rheumatism) Orphan Disease Programme. Additional funding from Scleroderma and Raynaud’s UK allowed a 1-year extension of the study.

Contributors: ALH, ML, RH, LM, AJS, EB, LaC, JHWD, OD, KF, WJG, RO, MCV and CPD were members of the European Scleroderma Observational Study (ESOS) Steering Committee and designed the ESOS study. SeP and ML were responsible for the statistical analysis. ALH, RH, LaC, JHWD, OD, MV, CoA, VHO, DF, MH, MM-C, AB-G, OM, PJ, ACJ, WS, PM, FCH, ChA, MEA, ED, RM, MA, MHB, LoC, NSD, HG, PL, YA, KC, SJ, AJM, NM, UM-L, GR, MB, JR, PEC, A-LF, EH, JH, MI, JSM, JMvL, SaP, SuP, AR, JS, BC, CS, TS, DJV, CG, G-ST and CPD were principal investigators at the different sites and recruited patients. XP and GD were study coordinators. ALH, SeP, ML, RH, LM, AJS and CPD wrote the draft report, and all authors reviewed the report and provided comments. Funding: ESOS was funded by a grant from the EULAR (European League Against Rheumatism) Orphan Disease Programme. Additional funding from Scleroderma and Raynaud’s UK allowed a 1-year extension of the study. Competing interests: ALH has done consultancy work for Actelion, served on a Data Safety Monitoring Board for Apricus, received research funding and speaker’s fees from Actelion, and speaker’s fees from GSK. JHWD has consultancy relationships and/or has received research funding from Actelion, BMS, Celgene, Bayer Pharma, Boehringer Ingelheim, JB Therapeutics, Sanofi-Aventis, Novartis, UCB, GSK, Array BioPharma, Active Biotech, Galapagos, Inventiva, Medac, Pfizer, Anamar and RuiYi, and is stock owner of 4D Science. OD has received consultancy fees from 4D Science, Actelion, Active Biotech, Bayer, Biogenidec, BMS, Boehringer Ingelheim, EpiPharm, Ergonex, espeRare Foundation, Genentech/Roche, GSK, Inventiva, Lilly, Medac, Medimmune, Pharmacyclics, Pfizer, Serodapharm, Sinoxa and UCB, and received research grants from Actelion, Bayer, Boehringer Ingelheim, Ergonex, Pfizer and Sanofi, and has a patent mir-29 for the treatment of systemic sclerosis licensed. WJG has received teaching fees from Pfizer. CA has served as a consultant for AbbVie, Pfizer, Roche, UCB, MSD, BMS and Novartis, and has received research funding and speaker fees from AbbVie, Pfizer, Roche, UCB, MSD, BMS and Novartis. FCH has received research funding from Actelion. MEA has undertaken advisory board work and received honoraria from Actelion, and received speaker’s fees from Bristol-Myers Squibb. NSD has done consultancy for AbbVie, Pfizer, Roche and MSD, and received speaker’s fees from AbbVie, Boehringer-Ingelheim, Pfizer, Richter Gedeon, Roche and MSD. HG has done consultancy work and received honoraria from Actelion. UM-L is funded in part by EUSTAR, EULAR and the European Community (Desscipher programme). JMvL has received honoraria from Eli Lilly, Pfizer, Roche, MSD and BMS. SP has received research grants from Actelion Pharmaceuticals Australia, Bayer, GlaxoSmithKline Australia and Pfizer, and speaker fees from Actelion. AR receives funding from AstraZeneca. CPD has done consultancy for GSK, Actelion, Bayer, Inventiva and Merck-Serono, received research grant funding from GSK, Actelion, CSL Behring and Inventiva, received speaker’s fees from Bayer and given trial advice to Merck-Serono.