Measurement properties of the FATIGUE-PRO in patients with SLE in real-world settings Authors: Caroline Gordon, Matthias Schneider, Gamal Chehab, Stefan Cano, Antoine Regnault, Christian Stach, Christine de la Loge, Sophie Cleanthous, Thomas Morel Published: 2026 ABSTRACT AbstractObjectiveFatigue is a common, multifaceted symptom in SLE which is difficult to treat. There has been a lack of broadly accepted SLE-specific assessment tools for fatigue. This study assessed the psychometric properties of the FATIGUE-PRO, a new patient-reported outcome measure, in patients with SLE in real-world clinical outpatient settings (Birmingham, UK, and Düsseldorf, Germany).MethodsPatients with SLE aged ≥18 years from cross-sectional, observational studies in Birmingham and Düsseldorf completed the three FATIGUE-PRO domains (physical fatigue, mental and cognitive fatigue and susceptibility to fatigue). Psychometric evaluation was completed using Rasch Measurement Theory and classical test theory methods. Association between the FATIGUE-PRO and other clinical outcome assessments was evaluated using Spearman’s rank order correlations.ResultsOverall, 200 patients were recruited from Birmingham and 234 from Düsseldorf. The FATIGUE-PRO demonstrated consistently adequate measurement properties. Items demonstrated high levels of targeting (82–95%) and the response scale functioned as intended for most items across all domains (82−100% ordered thresholds). Domains were cohesive overall, with minimal item misfit, and demonstrated little local dependency. Principal component analysis of residuals was supportive of unidimensionality of all three domains. All domain scores were reliable and demonstrated high internal consistency reliability (0.90−0.97). Moderate-to-high correlations, supportive of construct validity, were observed between the FATIGUE-PRO scores and other patient-reported outcomes, including Lupus Quality of Life. Low correlations were seen with outcome measures such as Brief Index of Lupus Damage and clinician-reported SLE Disease Activity Index-2000.ConclusionsThe FATIGUE-PRO was demonstrated to be fit-for-purpose in real-world clinical outpatient settings, with strong measurement properties, supporting its use in the assessment of fatigue in SLE.Keywords: Health-Related Quality Of Life; Lupus Erythematosus, Systemic; Patient Reported Outcome Measures; Fatigue; Quality of LIfe WHAT IS ALREADY KNOWN ON THIS TOPIC Fatigue is a common and debilitating symptom in SLE, yet no broadly accepted disease-specific instrument has been available to comprehensively assess it. Fatigue is a common and debilitating symptom in SLE, yet no broadly accepted disease-specific instrument has been available to comprehensively assess it. The FATIGUE-PRO is a new SLE-specific patient-reported outcome measure covering three domains (physical fatigue, mental and cognitive fatigue and susceptibility to fatigue), initially validated in a cohort of patients from a phase 2b clinical trial. The FATIGUE-PRO is a new SLE-specific patient-reported outcome measure covering three domains (physical fatigue, mental and cognitive fatigue and susceptibility to fatigue), initially validated in a cohort of patients from a phase 2b clinical trial. WHAT THIS STUDY ADDS In this study, the FATIGUE-PRO was demonstrated to be fit-for-purpose with strong measurement properties in real-world clinical outpatient settings across two different countries. In this study, the FATIGUE-PRO was demonstrated to be fit-for-purpose with strong measurement properties in real-world clinical outpatient settings across two different countries. HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY These findings support the use of FATIGUE-PRO in the assessment of fatigue in SLE. Future research is needed to validate the measure in larger and more diverse patient populations, to assess how responses to the FATIGUE-PRO vary over time and in response to treatment and to compare against additional measures of fatigue. These findings support the use of FATIGUE-PRO in the assessment of fatigue in SLE. Future research is needed to validate the measure in larger and more diverse patient populations, to assess how responses to the FATIGUE-PRO vary over time and in response to treatment and to compare against additional measures of fatigue. Introduction Fatigue is a highly prevalent and debilitating symptom in SLE, which significantly impacts patients’ health-related quality of life (HRQoL), including how they feel and their ability to function, learn, work and care for family.1–3 Despite recent advances in the understanding of fatigue,4 it remains a complex symptom which can be challenging to measure, as acknowledged by the US Food and Drug Administration (FDA) in its SLE guidance.4 5 There is also disagreement regarding the extent to which fatigue correlates with disease activity, with many studies suggesting low correlation between the two.6 7 Possibly due to challenges measuring fatigue, it has historically been difficult to evaluate and treat.4 5 Reliable methods of evaluation, in clinical practice and clinical trials, are important to facilitate assessment of the impact of treatment on fatigue.8 While general fatigue measures are widely used in SLE,9 10 development of SLE-specific measures has been limited. Existing measures, at the time of this study design, for assessing fatigue in SLE included the Functional Assessment of Chronic Illness Therapy-Fatigue (FACIT-Fatigue) scale, Fatigue Severity Scale (FSS) and Lupus Quality of Life (LupusQoL).2 9–12 FACIT-Fatigue is a scale developed from the Functional Assessment of Cancer Therapy-Anemia measure, originally developed for cancer and since used in many other conditions, and is validated for use in SLE.10–13 It includes 13 items assessing fatigue and its impact on patients’ daily lives.2 14 The FSS is a nine-item scale assessing the severity and impact of fatigue on physical, social and cognitive functions, used in chronic conditions like multiple sclerosis and SLE.9 12 LupusQoL is an SLE-specific instrument covering a range of concepts, including a four-item fatigue domain.11 Psychometric analyses of the FACIT-Fatigue scale and LupusQoL in SLE clinical trials have indicated some limitations related to the comprehensiveness of concept coverage and conceptual clarity for some items.14 15 These findings suggested that development of a patient-reported outcome (PRO) measure of fatigue for use in SLE clinical trials would require empirical exploration from the most detailed patient experience to the broadest theoretical levels of the concept of interest. The FDA SLE Guidance for Industry highlights that a measure with a clear definition of fatigue as it relates to patients with SLE, a clear conceptual framework describing fatigue and methods for measuring fatigue symptoms and effect in the presence of comorbid factors is needed.5 The different aspects of fatigue, including physical and mental components, are required to ensure adequate conceptual coverage.5 A previous in-depth conceptualisation of fatigue in SLE demonstrated the complexity and multidimensionality of the symptom experience,4 offering the basis for development of the FATIGUE-PRO to quantify this core SLE symptom. The FATIGUE-PRO is a new SLE-specific PRO measure aimed at capturing the multifaceted nature of fatigue.16 It comprises 31 items across three domains: physical fatigue, mental and cognitive fatigue, and susceptibility to fatigue. The FATIGUE-PRO was developed in line with FDA guidelines, using a mixed-methods psychometric approach.16 17 It benefitted from the comprehensive conceptualisation of the fatigue experience in SLE4 and was initially validated using patient data from the phase 2b RISE trial (NCT02804763) of dapirolizumab pegol in patients with moderate-to-severe, active SLE.18 Since the FATIGUE-PRO was previously evaluated in a cohort of patients from a clinical trial, this study aimed to assess the psychometric properties of the FATIGUE-PRO in patients with SLE in real-world clinical outpatient settings (Birmingham, UK, and Düsseldorf, Germany). We assessed the extent to which the FATIGUE-PRO measurement continuum is stable across two countries and languages and conducted an initial exploration into the different aspects of the fatigue experience, by reviewing the sequencing of the FATIGUE-PRO items across fatigue severity levels. Finally, we evaluated associations between the FATIGUE-PRO scores and other clinician-reported outcome and PRO assessments. Methods Patients and study design In Birmingham, UK, and Düsseldorf, Germany, ambulatory patients with SLE were prospectively recruited into cross-sectional, observational studies in rheumatology outpatient clinics. In Birmingham, the FATIGUE-PRO was included in an observational study in the dedicated lupus clinic at the Sandwell and West Birmingham NHS Trust. In Düsseldorf, the FATIGUE-PRO was included in an epidemiological study during routine clinical examinations at the Polyclinic and Functional Area for Rheumatology and Hiller Rheumatology Research Center. The FATIGUE-PRO was originally developed in US English, thus UK English and German versions were obtained by a linguistic validation process, prior to the current research, following established good practice principles.19 Patients in Birmingham were assessed between June 2017 and October 2018 and between September 2016 and December 2017 in Düsseldorf. Patients were required to be ≥18 years old with a confirmed SLE diagnosis. Patients in the Birmingham cohort were required to fulfil at least four of the 1997 revised American College of Rheumatology classification criteria for SLE at the time of study conduct.20 21 Patients in the Düsseldorf cohort required a rheumatologist-confirmed diagnosis of SLE with the International Classification of Diseases (ICD)-10 M32 code, also based on the 1997 revised American College of Rheumatology classification criteria.20 21 Proficiency in written English and German was required in the respective cohorts. The original development cohort included clinical trial patients with moderate to severe disease activity,16 while in the present study there were no inclusion criteria regarding disease activity levels. Patients were excluded from the Birmingham cohort if they had medical, language or educational reasons that would prevent them from completing the FATIGUE-PRO. Patients with a history of drug or alcohol abuse in the previous 12 months, patients who were pregnant and patients who were already included in interventional clinical trials were also excluded in Birmingham. Patients were required to provide written consent. The studies were approved by the relevant local ethics committees: the North East–Newcastle North Tyneside Research Ethics Committee (16/NE/0414) and the Ethics Committee of the Medical Faculty of the Heinrich Heine University Düsseldorf (5466R). Patient and public involvement The FATIGUE-PRO was originally developed in collaboration with patients16; however, patients and/or the public were not involved in the design, conduct or reporting of this study. Outcomes The 31 items of the FATIGUE-PRO cover three domains: physical fatigue, mental and cognitive fatigue and susceptibility to fatigue (online supplemental table S1).16 For each item, a five-point Likert scale is used to quantify frequency of symptoms over the past seven days. For each domain, the sum of item scores is linearly transformed to provide a score from 0 to 100, with higher scores indicating higher levels of fatigue.16 lupus-13-2-s001.pdf In both studies, other clinician-reported outcomes that were recorded included the SLE Disease Activity Index-2000 (SLEDAI-2K; score range: 0–105, lower scores reflect less disease activity)22 23 and Systemic Lupus International Collaborating Clinics/American College of Rheumatology Damage Index (SDI; score range: 0–47, lower scores reflect less organ damage).24 25 In the Birmingham cohort, the British Isles Lupus Assessment Group (BILAG)-2004 index of disease activity (numerical score calculated according to Yee et al; domain or system score range: 0–12, lower scores reflect less disease activity),26 27 and the LupusQoL (domain score range: 0–100, higher scores reflect better HRQoL) were also completed.11 In the Düsseldorf cohort, other PROs included the FSS (total score range: 9–63, lower scores reflect better HRQoL),9 12 Systemic Lupus Activity Questionnaire (SLAQ; total score range: 0–44, lower scores reflect less disease activity),22 28 Brief Index of Lupus Damage (BILD; total score range: 0–31, lower scores reflect less organ damage)22 29 and Hannover Functional Ability Questionnaire (FFbH; score range: 0−100%, lower percentages reflect greater limitation in performing daily activities)30 31; FFbH scores were converted to Health Assessment Questionnaire (HAQ; score range: 0–3, lower scores reflect better health) scores.32 33 Analytical techniques and statistical analysis Rasch Measurement Theory (RMT) examined whether item response data achieved the requirements specified by the measurement model in relation to various tests: overall fit of the data to the Rasch model, scale-to-sample targeting, item response thresholds (suitability of response scale), item performance fit, assessment of unidimensionality (through principal component analysis [PCA] of residuals from the Rasch model), local dependence and item stability. Classical test theory (CTT) was also applied to complement RMT psychometric evaluation. Internal consistency reliability was assessed using Person separation index (PSI) and Cronbach’s alpha. Spearman’s rank order correlations were calculated to assess construct validity through associations between the FATIGUE-PRO domain scores and BILAG-2004, SLEDAI-2K, SDI, LupusQoL, FSS, SLAQ, BILD, FFbH and HAQ scores. Further details of RMT and CTT methods are available in online supplemental table S2. RMT analyses were conducted using RUMM2030 software (RUMM Laboratory, Perth, Australia). Other analyses were conducted using SAS V.9.4 (SAS Institute, Cary, North Carolina, USA). No formal statistical comparisons were performed due to the exploratory nature of the analyses. Results Baseline demographics and disease characteristics Baseline demographics and disease characteristics of the patients in the Birmingham (n=200) and Düsseldorf (n=234) cohorts are shown in table 1. Approximately half of the patients were 40–60 years old (Birmingham: 51.5% [103/200]; Düsseldorf: 47.9% [112/234]). In the Birmingham cohort, there was a greater diversity of race; 51.0% (102/200) of patients were white/Caucasian, compared with 99.1% (232/234) in the Düsseldorf cohort. Patients were predominantly female (Birmingham: 92.5% [185/200]; Düsseldorf: 87.6% [205/234]). The mean (SD) number of years since SLE diagnosis was 15.8 (9.3) years in the Birmingham cohort, compared with 17.6 (9.5) years in the Düsseldorf cohort. Patients in Birmingham exhibited lower disease activity and organ damage. Mean (SD) SLEDAI-2K was 2.4 (3.0) in the Birmingham cohort, compared with 4.0 (4.9) in the Düsseldorf cohort. Mean (SD) SDI was 1.3 (1.6) in the Birmingham cohort and 1.6 (2.0) in the Düsseldorf cohort. Within both cohorts, patients received various SLE treatments; the full list of treatments is provided in table 1. Table 1Baseline demographics and disease characteristicsBirminghamn=200Düsseldorfn=234Sex, n (%) Female185 (92.5)205 (87.6) Male15 (7.5)29 (12.4)Age, years, n (%) <203 (1.5)5 (2.1) 20–4062 (31.0)87 (37.2) 40–60103 (51.5)112 (47.9) >6032 (16.0)30 (12.8)Race, n (%) White/Caucasian102 (51.0)232 (99.1) Black42 (21.0)2 (0.9) Other38 (19.0)0 (0.0) South Asian18 (9.0)0 (0.0)Time since SLE diagnosis, years Mean (SD)15.8 (9.3)17.6 (9.5) Range1–481–44BILAG-2004* Mean (SD)3.0 (39.0)N/A† Range0−26N/A†SLEDAI-2K (disease activity) Mean (SD)2.4 (3.0)4.0 (4.9) Range0–160–29SDI (organ damage) Mean (SD)1.3 (1.6)1.6 (2.0) Range0–80–11SLE treatments,‡ n (%)Antimalarials Hydroxychloroquine127 (63.5)118 (50.4) Chloroquine1 (0.5)35 (15.0) Mepacrine2 (1.0)0 (0.0)Glucocorticoids/steroids Oral glucocorticoids117 (58.5)129 (55.1) Intralesional/intramuscular steroids2 (1.0)0 (0.0) Intravenous steroids1 (0.5)0 (0.0) Topical steroids1 (0.5)0 (0.0)Immunosuppressants/calcineurin inhibitors Mycophenolate mofetil55 (27.5)34 (14.5) Azathioprine26 (13.0)36 (15.4) Ciclosporin A8 (4.0)10 (4.3) Tacrolimus2 (1.0)2 (0.9) Biologic8 (4.0)27 (11.5) Methotrexate6 (3.0)9 (3.8) Leflunomide1 (0.5)0 (0.0) Cyclophosphamide0 (0.0)2 (0.9)*Numerical score was calculated according to Yee et al.27†BILAG-2004 scores were not collected in the Düsseldorf cohort.‡Total percentage >100% because patients could have been treated with several SLE treatments at a time.BILAG-2004, British Isles Lupus Assessment Group-2004; SDI, Systemic Lupus International Collaborating Clinics/American College of Rheumatology Damage Index; SLE, systemic lupus erythematosus; SLEDAI-2K, SLE Disease Activity Index-2K. Baseline demographics and disease characteristics *Numerical score was calculated according to Yee et al.27 †BILAG-2004 scores were not collected in the Düsseldorf cohort. ‡Total percentage >100% because patients could have been treated with several SLE treatments at a time. BILAG-2004, British Isles Lupus Assessment Group-2004; SDI, Systemic Lupus International Collaborating Clinics/American College of Rheumatology Damage Index; SLE, systemic lupus erythematosus; SLEDAI-2K, SLE Disease Activity Index-2K. Psychometric evaluation of FATIGUE-PRO measurement properties The FATIGUE-PRO demonstrated consistently adequate measurement properties within all three domains in both cohorts (table 2; figure 1), shown through RMT, scale-to-sample targeting and threshold mapping results. Overall, the item sets for the three domains fit the Rasch model (non-significant χ2, p>0.05; table 2) in both cohorts, suggesting that they overall met the requirements of the Rasch model. Table 2Summary of FATIGUE-PRO measurement propertiesMeasurement propertyPhysical fatigue (nine items)Mental and cognitive fatigue (11 items)Susceptibility to fatigue (11 items)BirminghamDüsseldorfBirminghamDüsseldorfBirminghamDüsseldorfOverall fit of the data to the Rasch model (χ2, p value): do the observed responses of patients to the items deviate from the Rasch model?*72.39 (p=0.74)95.17 (p=0.13)103.18 (p=0.37)103.60 (p=0.36)89.98 (p=0.73)93.91 (p=0.63)RMT targeting: how adequate is the scale-to-sample targeting?†86%91%82%85%85%95%Response thresholds: proportion of item response thresholds working as intended‡100%100%100%100%91%§82%¶RMT item fit: proportion of items with acceptable fit based on residuals and the χ2 statistic Fit residuals**100%78%††91%‡‡82%§§100%91%¶¶ χ2***100%89%†††100%91%‡‡‡100%100%RMT item dependency: proportion of item pairs that are locally independent§§§97%¶¶¶97%****98%††††98%‡‡‡‡96%§§§§95%¶¶¶¶RMT PSI: are patients in the sample separated by the domain items?*****0.950.920.950.920.940.89Cronbach’s alpha: a measure of internal consistency†††††0.960.940.970.950.960.91Higher percentages indicate better findings.*Non-significant χ2 (p>0.05) indicates no deviation of the observed data from the theoretical model, that is, good fit.†Estimated using the percentage of individual sample measurements (n=916) covered by the domain range; higher percentages indicate better targeting.‡Ideally, 100% of thresholds should be ordered.§‘Feel like sleeping all day long?’ demonstrated disordered response thresholds.¶‘Need to lie down during the day?’ and ‘Feel like sleeping all day long?’ demonstrated disordered response thresholds.**Percentage of items within the recommended range of −2.5 to 2.5.††‘Feel a lack of energy?’ and ‘Feel that it was physically hard to get up and start moving?’ demonstrated item misfit.‡‡‘Feel mentally tired?’ demonstrated item misfit.§§'Feel that it was hard to concentrate?’ and ‘Feel like it was difficult to be motivated?’ demonstrated item misfit.¶¶‘Not feel like getting out of bed in the morning?’ demonstrated item misfit.***Percentage of items with a non-significant χ2 probability.†††‘Feel that it was physically hard to get up and start moving?’ demonstrated item misfit.‡‡‡‘Feel like it was difficult to be motivated?’ demonstrated item misfit.§§§Dependency was defined as the number of item pairs that were locally dependent on the basis of >0.3 residual correlations indicating >9% shared variance.¶¶¶‘Feel a lack of energy?’ and ‘Feel physically tired?’ demonstrated dependency.****'Feel that your arms were weak?’ and ‘Feel that your arms were heavy?’ demonstrated dependency.††††‘Feel you could not remember things?’ and ‘Feel like you had “brain fog”?’ demonstrated dependency.‡‡‡‡‘Feel that it was hard to concentrate?’ and ‘Feel that it was hard to stay focused for a long period of time?’ demonstrated dependency.§§§§‘Wake up feeling tired?’ and ‘Feel that you did not get the amount of sleep you needed?’, and ‘Need to lie down during the day?’ and ‘Feel the need to sleep during the day?’ demonstrated dependency.¶¶¶¶‘Wake up feeling tired?’ and ‘Feel you did not get the amount of sleep you needed?’, ‘Feel it was hard to maintain activities that required physical effort for long periods?’ and ‘Feel it was hard to maintain activities that required mental effort for long periods?’ and ‘Feel it was hard to maintain activities that required physical effort for long periods?’ and ‘Feel you needed to take breaks from activities?’ demonstrated dependency.*****PSI was reported on a scale from 0 to 1; 0=all error; 1=no error; a coefficient of ≥0.70 was considered to represent acceptable internal consistency reliability.†††††Cronbach’s alpha was reported on a scale from 0 to 1; 0=no internal consistency; 1=perfect internal consistency; a coefficient of ≥0.70 was considered to represent acceptable internal consistency reliability.PSI, Person Separation Index; RMT, Rasch Measurement Theory. Summary of FATIGUE-PRO measurement properties Higher percentages indicate better findings. *Non-significant χ2 (p>0.05) indicates no deviation of the observed data from the theoretical model, that is, good fit. †Estimated using the percentage of individual sample measurements (n=916) covered by the domain range; higher percentages indicate better targeting. ‡Ideally, 100% of thresholds should be ordered. §‘Feel like sleeping all day long?’ demonstrated disordered response thresholds. ¶‘Need to lie down during the day?’ and ‘Feel like sleeping all day long?’ demonstrated disordered response thresholds. **Percentage of items within the recommended range of −2.5 to 2.5. ††‘Feel a lack of energy?’ and ‘Feel that it was physically hard to get up and start moving?’ demonstrated item misfit. ‡‡‘Feel mentally tired?’ demonstrated item misfit. §§'Feel that it was hard to concentrate?’ and ‘Feel like it was difficult to be motivated?’ demonstrated item misfit. ¶¶‘Not feel like getting out of bed in the morning?’ demonstrated item misfit. ***Percentage of items with a non-significant χ2 probability. †††‘Feel that it was physically hard to get up and start moving?’ demonstrated item misfit. ‡‡‡‘Feel like it was difficult to be motivated?’ demonstrated item misfit. §§§Dependency was defined as the number of item pairs that were locally dependent on the basis of >0.3 residual correlations indicating >9% shared variance. ¶¶¶‘Feel a lack of energy?’ and ‘Feel physically tired?’ demonstrated dependency. ****'Feel that your arms were weak?’ and ‘Feel that your arms were heavy?’ demonstrated dependency. ††††‘Feel you could not remember things?’ and ‘Feel like you had “brain fog”?’ demonstrated dependency. ‡‡‡‡‘Feel that it was hard to concentrate?’ and ‘Feel that it was hard to stay focused for a long period of time?’ demonstrated dependency. §§§§‘Wake up feeling tired?’ and ‘Feel that you did not get the amount of sleep you needed?’, and ‘Need to lie down during the day?’ and ‘Feel the need to sleep during the day?’ demonstrated dependency. ¶¶¶¶‘Wake up feeling tired?’ and ‘Feel you did not get the amount of sleep you needed?’, ‘Feel it was hard to maintain activities that required physical effort for long periods?’ and ‘Feel it was hard to maintain activities that required mental effort for long periods?’ and ‘Feel it was hard to maintain activities that required physical effort for long periods?’ and ‘Feel you needed to take breaks from activities?’ demonstrated dependency. *****PSI was reported on a scale from 0 to 1; 0=all error; 1=no error; a coefficient of ≥0.70 was considered to represent acceptable internal consistency reliability. †††††Cronbach’s alpha was reported on a scale from 0 to 1; 0=no internal consistency; 1=perfect internal consistency; a coefficient of ≥0.70 was considered to represent acceptable internal consistency reliability. PSI, Person Separation Index; RMT, Rasch Measurement Theory. Scale-to-sample targeting and threshold mapping of the three FATIGUE-PRO domains. Sample distributions (grey bars representing total score distributions for each FATIGUE-PRO domain) and item threshold distributions (yellow, green and blue bars) plotted on the same linear measurement continuum of fatigue severity, with lower scores indicating lower levels of fatigue. Item threshold maps show the most probable response category across the measurement continuum. Targeting is assessed by examining the relative range and coverage of the sample distribution by item thresholds. Items are ordered top to bottom by increasing difficulty. RMT expects ordering of response categories to reflect intended severity. RMT, Rasch Measurement Theory. Unidimensionality of the FATIGUE-PRO was confirmed across all domains through PCA of residuals from the Rasch model in the Birmingham and Düsseldorf cohorts separately; no consistent and clear underlying structure was found in Rasch model residuals for any of the domains in the two cohorts, suggesting that once the common source of variation captured by the Rasch model was accounted for, the responses to the items were not consistently and strongly influenced by another factor (online supplemental tables S3 and S4). In the physical fatigue domain, items demonstrated high levels of targeting, with item thresholds covering 86% of sample measurements in the Birmingham cohort and 91% in the Düsseldorf cohort, suggesting accurate measurement of physical fatigue (table 2). The item response thresholds were ordered successively, as expected, in 100% of cases in both cohorts, indicating the response options worked as intended for all physical fatigue items (table 2; figure 1A). Item fit statistics did not indicate significant item misfit (Birmingham χ2: 100%; Düsseldorf χ2: 89%), suggesting that the item set was cohesive overall. In the Düsseldorf cohort, some misfit was seen in the ‘Feel that it was physically hard to get up and start moving?’ item; further details on item fit statistics are available in online supplemental table S5. Most items were locally independent of each other, 97% in both cohorts, meaning they measured different aspects of physical fatigue clearly (table 2). In the mental and cognitive fatigue domain, targeting results were very good, with item thresholds covering 82% of sample measurements in the Birmingham cohort and 85% in the Düsseldorf cohort (table 2). Item response thresholds were again ordered successively, as expected, in 100% of cases in both cohorts, indicating the response options worked as intended for all domain items (table 2; figure 1B). Overall, the domain demonstrated an item fit χ2 value of 100% for the Birmingham cohort and 91% for the Düsseldorf cohort, suggesting that it was cohesive overall, except for one item in the Düsseldorf cohort (‘Feel like it was difficult to be motivated?’; table 2; online supplemental table S5). Similar to the physical fatigue domain, the items demonstrated minimal local dependency, 98% in both cohorts (table 2). In the susceptibility to fatigue domain, targeting was very good in the Birmingham cohort (85%) and excellent in the Düsseldorf cohort (95%; table 2). Results for the item response thresholds were lower compared with the other domains (Birmingham: 91%; Düsseldorf: 82%; table 2; figure 1C). Potential issues with the cohesiveness of the domain were observed in two items with disordered item thresholds (‘Need to lie down during the day?’ and ‘Feel like sleeping all day long?’); however, the domain was cohesive overall, and items otherwise demonstrated little local dependency (Birmingham: 96%; Düsseldorf: 95%; table 2). Across all domains, and both cohorts, PSI levels were high, denoting high levels of reliability (table 2). Internal consistency, assessed by Cronbach’s alpha, was high across all three domains (0.91−0.97), reflecting that items in each domain were measuring the same underlying concept (table 2). The ordering of items is presented in figure 1. Patterns demonstrated across the three domains aligned with the original validation of the measure, whereby consistencies in the items relevant at lower or higher levels of fatigue were observed. Inconsistencies, and thus lower stability, were observed in nine items between the two cohorts, through a uniform differential item functioning (DIF; p<0.01 with Bonferroni correction): ‘Feel a lack of energy?’, ‘Feel that your arms were weak?’, ‘Feel that your whole body was weak?’, ‘Feel that your whole body was heavy?’, ‘Feel like you had a “brain fog”?’, ‘Feel that it was hard to stay focused for a long period of time?’, ‘Wake up feeling tired?’, ‘Suddenly tired for no apparent reason?’ and ‘Feel the need to sleep during the day?’ (online supplemental table S6). The relative location of each of the 31 items, plotted on the continuum of fatigue, is shown in figure 2. These data suggest that lower levels of fatigue manifest in physical fatigue and susceptibility to fatigue, while mental and cognitive fatigue manifests at higher levels of fatigue. FATIGUE-PRO relative item mean locations. Mean location of each of the 31 items of the FATIGUE-PRO, plotted on the continuum of fatigue severity (across the Birmingham and Düsseldorf studies). Items at either end of the continuum are displayed for each domain of the FATIGUE-PRO. Data suggest that lower levels of fatigue severity manifest in physical fatigue and susceptibility to fatigue, while mental and cognitive fatigue manifests at higher levels of fatigue. FATIGUE-PRO association with other clinical outcome assessments Spearman’s correlation coefficients between the FATIGUE-PRO and clinician-reported outcomes (BILAG-2004, SLEDAI-2K and SDI), to assess the construct validity of the FATIGUE-PRO scores, are displayed in table 3. Overall, correlation coefficients <0.2 were observed between FATIGUE-PRO and all clinical measures, suggesting limited association between clinical measures of disease activity and patient experience of fatigue. Table 3Spearman’s rank order correlations between FATIGUE-PRO scores and clinical outcomes measuring disease activity or damage and other PRO scoresPhysical fatigueMental and cognitive fatigueSusceptibility to fatigueDisease activity and damage outcomes BILAG-2004 Birmingham0.020.050.06 SLEDAI-2K Birmingham−0.03−0.030.01 Düsseldorf0.130.120.08 SDI Birmingham0.150.030.05 Düsseldorf0.190.150.14PRO instruments LupusQoL domains Physical health−0.67−0.62−0.72 Pain−0.71−0.67−0.76 Planning−0.56−0.51−0.65 Intimate relationships−0.52−0.50−0.56 Burden to others−0.58−0.58−0.61 Emotional health−0.56−0.58−0.56 Body image−0.64−0.52−0.61 Fatigue−0.69−0.77−0.83 FSS0.610.600.60 SLAQ0.600.490.58 BILD0.120.150.11 FFbH−0.64−0.45−0.46 HAQ0.640.450.46Moderate correlations (0.30≤|r|<0.70) shown in light blue and high correlations (|r|≥0.70) shown in light green.For the BILAG-2004, SLEDAI-2K, SDI, FSS, SLAQ, BILD and HAQ instruments, a lower score indicates improvement. For the LupusQoL and FFbH, a higher score indicates improvements.BILAG-2004, British Isles Lupus Assessment Group-2004; BILD, Brief Index of Lupus Damage; FFbH, Funktionsfragebogen Hannover; FSS, Fatigue Severity Scale; HAQ, Health Assessment Questionnaire; LupusQoL, Lupus Quality of Life; PRO, patient-reported outcome; SDI, Systemic Lupus International Collaborating Clinics/American College of Rheumatology Damage Index; SLAQ, Systemic Lupus Activity Questionnaire; SLEDAI-2K, SLE Disease Activity Index-2K. Spearman’s rank order correlations between FATIGUE-PRO scores and clinical outcomes measuring disease activity or damage and other PRO scores Moderate correlations (0.30≤|r|<0.70) shown in light blue and high correlations (|r|≥0.70) shown in light green. For the BILAG-2004, SLEDAI-2K, SDI, FSS, SLAQ, BILD and HAQ instruments, a lower score indicates improvement. For the LupusQoL and FFbH, a higher score indicates improvements. BILAG-2004, British Isles Lupus Assessment Group-2004; BILD, Brief Index of Lupus Damage; FFbH, Funktionsfragebogen Hannover; FSS, Fatigue Severity Scale; HAQ, Health Assessment Questionnaire; LupusQoL, Lupus Quality of Life; PRO, patient-reported outcome; SDI, Systemic Lupus International Collaborating Clinics/American College of Rheumatology Damage Index; SLAQ, Systemic Lupus Activity Questionnaire; SLEDAI-2K, SLE Disease Activity Index-2K. Spearman’s correlation coefficients between the FATIGUE-PRO and other PRO scores are presented in table 3. FATIGUE-PRO showed moderate (0.30≤|r|<0.70) or high (|r|≥0.70) correlation with all PRO measures assessing similar or related concepts (LupusQoL, FSS, SLAQ, FFbH and HAQ). The highest correlations observed were between the FATIGUE-PRO and LupusQoL, particularly the LupusQoL fatigue and pain domains. Low correlations were observed between the FATIGUE-PRO and BILD, a patient-reported measure of damage in SLE. Discussion In this study, the psychometric properties of the recently developed FATIGUE-PRO were assessed in clinical practice, with results supporting the use of the measure in quantifying fatigue in patients with SLE in clinical outpatient settings. The FATIGUE-PRO was demonstrated to be fit-for-purpose and consistently satisfied stringent psychometric criteria across two cohorts with differing cultural and language backgrounds, reinforcing it as a useful, reliable and valid PRO measure of fatigue. The FATIGUE-PRO was evaluated in two distinct ambulatory patient cohorts, with English and German as their primary languages. This differs from the patient population included in the original FATIGUE-PRO development, who were mostly English-speaking and Spanish-speaking patients enrolled in the phase 2b RISE trial.16 18 Patients from the RISE trial presented with more severe disease activity, with greater SLEDAI-2K scores than in the current study, although with a shorter disease duration.16 Despite differences in the patient populations and context of use (real-world clinical practice vs clinical trial), the FATIGUE-PRO psychometric properties reported here were generally consistent with those obtained in patients from the RISE trial.16 Internal consistency reliability assessed by PSI was high across all domains, indicating that the measure can reliably distinguish patients with different levels of fatigue. The PSI range in the original RISE trial cohort was 0.92–0.94,16 consistent with the 0.89–0.95 range observed in these real-world cohorts here. Internal consistency assessed by Cronbach’s alpha was 0.95–0.96 across the FATIGUE-PRO domains in the original RISE cohort,16 aligning with the range of values observed here (0.90–0.97). By assessing DIF between two separate countries with the measure completed in two different languages, the invariance of the FATIGUE-PRO across both languages and cultures, and the extent of consistency across items between subgroups, could be evaluated. DIF findings in the present study indicated that items were largely consistent across the two cohorts and that the FATIGUE-PRO reliably captured the patient experience of fatigue across these groups. However, nine items presented some DIF across the two cohorts. The identified DIF observed may have arisen from linguistic or item inequivalence.34 Subtle differences in meaning between the original English items and their German translations may have influenced how patients interpreted certain questions, particularly for colloquial expressions such as ‘brain fog’, perhaps as a product of cultural differences. It should be noted, however, that the German version was obtained by a linguistic validation process following established good practice principles,19 with efforts made to ensure cultural equivalence between the original and translated versions. The relatively small number of patients in the cohorts here could possibly have also contributed to the observed DIF. Additionally, it is possible that the experience of fatigue between patients in the UK and Germany is perceived or expressed differently, resulting in varied responses. Further studies in larger cohorts of patients are required to assess this and the implications for other language translations. The item ordering of all 31 items of the FATIGUE-PRO across the two cohorts suggests that symptoms of physical fatigue and susceptibility to fatigue tend to manifest at lower levels of fatigue, while symptoms of mental and cognitive fatigue tend to manifest at higher levels of fatigue. This is consistent with feedback obtained from patient focus groups held during the original development and validation of the FATIGUE-PRO:“…physical fatigue it’s kind of always there but when you are really tired your brain goes too… that’s when the mental fatigue kicks in” .16 In line with expectations and previous literature, the FATIGUE-PRO demonstrated stronger associations with other PROs versus clinician-reported measures of disease activity and damage.14 16 In the original development cohort, low correlation with SLEDAI-2K was also observed.16 Similarly, in a psychometric validation of the FACIT-Fatigue scale across the BLISS-SC, BLISS-52 and BLISS-76 clinical trials, moderate-to-high correlation was observed with the Short Form-36 Health survey, a PRO measure of HRQoL, compared with low-to-moderate correlation with BILAG-2004.14 The role of disease activity in fatigue in SLE has long been debated and varies by disease activity measure.6 35 When assessed with the Systemic Lupus Activity Measure, which includes a question on fatigue, significant correlation between fatigue and disease activity has been observed,35 compared with low correlation when assessed with SLEDAI.36 The data presented here suggest weak associations between fatigue and disease activity, despite patients with a range of disease activity levels being included; further research is required to provide additional evidence, particularly for patients with severe disease activity. Inclusion of fatigue-related items in disease activity measures has also been debated; due to this potential lack of agreement, fatigue items may not be included in novel composite endpoints of disease activity.37 This highlights the need for independent SLE-specific measures which accurately and fully evaluate fatigue in SLE, where general fatigue measures may not capture comprehensively the patient experience of fatigue in SLE. The LupusQoL was developed with significant patient involvement at all stages, similarly to the FATIGUE-PRO.11 Associations between the FATIGUE-PRO and LupusQoL domains were as expected, with high correlations observed between the FATIGUE-PRO and the LupusQoL fatigue and pain domains. This is in line with previous research on the relation between pain and fatigue.38–40 Further research including greater numbers of patients is needed to assess how the three FATIGUE-PRO domains vary in comparison with the LupusQoL domains over time. The FSS is a nine-item PRO assessing the physical, social and cognitive effects of fatigue,9 12 and a moderate correlation was observed between the FATIGUE-PRO and FSS. This suggests that additional information may be covered by the FATIGUE-PRO; while the FSS covers items relating to the three FATIGUE-PRO domains, these are not covered in the same breadth and depth, reinforcing the FATIGUE-PRO as a valuable, unique measure in evaluating fatigue in SLE. Additionally, while the FSS and FATIGUE-PRO have the same recall period, the FSS asks patients to rank the severity of symptoms, while the FATIGUE-PRO asks patients to assess the frequency of their symptoms, which may also contribute to the moderate association. Low correlations between the FATIGUE-PRO and BILD may have been expected given the low correlations noted above between the FATIGUE-PRO and clinical measures of disease activity and damage. A strength of this study was the comparison between two distinct cohorts, with varied language, cultural and ethnic backgrounds. Despite the different settings, the FATIGUE-PRO demonstrated broadly consistent properties across the cohorts, providing supportive evidence of its adequate psychometric properties and building on previous evidence from clinical trial data. Limitations of this research include the lack of longitudinal data to assess sensitivity to change and test-retest reliability of the FATIGUE-PRO scores and lack of definition of meaningful within-patient change. Nevertheless, results from this study are overall supportive and call for further documentation of longitudinal measurement properties of the FATIGUE-PRO in future complementary research. As patients were recruited from single sites in Birmingham and Düsseldorf, it should also be noted that the cohorts here may not be fully representative of the broader SLE patient population. Finally, fatigue in patients with SLE is multifactorial, and other associated conditions such as fibromyalgia may also contribute to fatigue41; however, these were not assessed formally as part of this study. In conclusion, despite the use of existing instruments, there remains a need for new, patient-centric, specific measures which comprehensively assess fatigue in SLE. This study demonstrated that the FATIGUE-PRO is a valuable measure to quantify fatigue in patients with SLE in clinical outpatient settings, as shown across two separate cultural and language backgrounds. Overall, the FATIGUE-PRO consistently satisfied stringent psychometric criteria, supporting it as a reliable PRO measure of fatigue, as it did in previous analyses of phase 2b clinical trial data.16 Some items presented differential functioning between the English and German cohorts; these findings are preliminary and they should be consolidated by more evidence, not only from English and German samples but extending to other cultures, to support use of the FATIGUE-PRO in a multicultural setting. Further research is also needed to assess how responses to the FATIGUE-PRO vary over time and in response to treatment. This, alongside validation in a larger cohort of clinical trial patients, is being investigated using data from the phase 3 PHOENYCS GO trial (NCT04294667) of dapirolizumab pegol in SLE.42 An evaluation of the potential for a FATIGUE-PRO total score and a ‘dual-scoring system’ is also required, where a total score may provide an assessment of overall fatigue, while the individual domains still provide a granular description of the patient experience. It is important for future research to document the psychometric properties of the FATIGUE-PRO in larger cohorts of patients not involved in clinical trials, such that a broader spectrum of disease activity can be considered, and to demonstrate the sensitivity to change of the three domains. Acknowledgments The authors thank the patients, the investigators and their teams who took part in both studies. The authors thank Sophia Tosounidou, MD, Sandwell and West Birmingham NHS Trust for patient assessments and local study management of the Birmingham cohort. The authors also acknowledge Juliette Meunier, Modus Outcomes, Lyon, France for their involvement in the studies. The authors also thank Heather Edens, PhD, UCB, Smyrna, Georgia, USA, and Valerie Zediak, PhD, Biogen, Cambridge, Massachusetts, USA for editorial review during manuscript development and publication coordination and Rebecca Light, BSc, Costello Medical, Manchester, UK, and Patrick Cox, BSc, Costello Medical, London, UK, for medical writing and editorial assistance based on the authors’ input and direction. This study was funded by UCB. Footnotes Contributors: TM is the guarantor. CG was involved in conceptualisation, data curation, funding acquisition, investigation, methodology, project administration, resources, supervision, validation and writing—review and editing; MS was involved in conceptualisation, data curation, investigation, methodology, project administration, resources, supervision, validation and writing—review and editing; GC was involved in data curation, investigation, resources and writing—review and editing; SCa was involved in conceptualisation, methodology, supervision and writing—review and editing; AR was involved in conceptualisation, formal analysis, methodology and writing—review and editing; CS was involved in data curation, funding acquisition, investigation, supervision and writing—review and editing; CdlL was involved in formal analysis and writing—review and editing; SCl was involved in conceptualisation, formal analysis, methodology, project administration, visualisation, writing—original draft and writing—review and editing; TM was involved in conceptualisation, formal analysis, investigation, methodology and writing—review and editing. Funding: This study was funded by UCB, and development of this manuscript was funded by UCB and Biogen. Support for third-party writing assistance for this article was funded by UCB and Biogen in accordance with Good Publication Practice 2022 guidelines (https://www.ismpp.org/gpp-2022). Competing interests: CG received consultancy fees from Alumis, Amgen, AstraZeneca and UCB. MS: consultant for AbbVie, AstraZeneca, BMS, GSK, Novartis, Otsuka and Roche; speaker’s bureau for AstraZeneca and GSK; received grant/research support from AstraZeneca and GSK. GC: received consultancy fees from AstraZeneca and GSK; received grant/research support from AstraZeneca and GSK. SCa, AR and SCl: employees of Modus Outcomes, a THREAD Company. CS: employee and shareholder of UCB. CdlL: consultant to UCB. TM: employee and shareholder of UCB. Patient and public involvement: Patients and/or the public were not involved in the design, conduct, reporting or dissemination plans of this research. Provenance and peer review: Not commissioned; externally peer reviewed. Supplemental material: This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise. Data availability statement Data are available upon reasonable request. The data underlying this article cannot be shared publicly due to the ethical approval of the study not permitting wider sharing of the data. Anonymised data may be shared on reasonable request to the corresponding author (Thomas.Morel@ucb.com), with the exception of demographic information, and will be evaluated on a case-by-case basis. Ethics statements Patient consent for publication Not applicable. Ethics approval This study involves human participants and was approved by the North East–Newcastle North Tyneside Research Ethics Committee (16/NE/0414) and the Ethics Committee of the Medical Faculty of the Heinrich Heine University Düsseldorf (5466R). Participants gave informed consent to participate in the study before taking part. References Associated Data Supplementary Materials lupus-13-2-s001.pdf Data Availability Statement Data are available upon reasonable request. The data underlying this article cannot be shared publicly due to the ethical approval of the study not permitting wider sharing of the data. Anonymised data may be shared on reasonable request to the corresponding author (Thomas.Morel@ucb.com), with the exception of demographic information, and will be evaluated on a case-by-case basis.