You are here

Article Text

Article menu

Download PDFPDF

Cutaneous lupus
Original research
Robust measurement of clinical improvement in patients with cutaneous lupus erythematosus
  1. Benjamin A Nanes,
  2. Jane L Zhu and
  3. Benjamin F Chong
  1. Department of Dermatology, University of Texas Southwestern Medical Center, Dallas, Texas, USA
  1. Correspondence to Dr Benjamin F Chong; Ben.Chong{at}UTSouthwestern.edu

Abstract

Objective The severity and disease course of cutaneous lupus erythematosus (CLE) are highly variable. Consequently, outcome measures for CLE clinical improvement are heterogeneous, complicating treatment decisions and therapeutic development. This study characterises CLE outcome measures and identifies the influence of clinical improvement thresholds on strengths of associations with patient demographic and clinical factors.

Methods In this pilot cohort study, multivariable models identified factors associated with CLE activity and skin damage improvement, defined as relative decreases in Cutaneous Lupus Activity and Severity Index (CLASI) activity (CLASI-A) and damage (CLASI-D) scores, over ranges of response thresholds.

Results 66 patients with 119 visit-pairs were included in the CLASI-A analysis. 74 patients with 177 visit-pairs were included in the CLASI-D analysis. Factors associated with CLE activity and damage improvement depended on the response threshold. Some associations were stronger at more stringent thresholds, including subacute CLE predominance with increased likelihood of CLASI-A improvement (R 2=0.73; 50% reduction: OR 1.724 (95% CI 0.537 to 5.536); 75%: 5.67 (95% CI 1.56 to 20.5)) and African-American race with decreased likelihood of CLASI-D improvement (R 2=0.80; 20%: 0.40 (95% CI 0.17 to 0.93); 40%: 0.25 (95% CI 0.08 to 0.82)). Other associations were stable across multiple thresholds, including older age of CLE development with increased likelihood of CLASI-A improvement (R 2=0.25; 50%: 1.05 (95% CI 1.01 to 1.09]; 75%: 1.05 (95% CI 1.00 to 1.10)) and higher initial disease activity with decreased likelihood of CLASI-D improvement (R 2=0.55; 20%: 0.91 (95% CI 0.84 to 0.98); 40%: 0.88 (95% CI 0.79 to 0.97)).

Conclusions Examining a range of CLASI threshold outcomes can comprehensively characterise changes in disease course in patients with CLE. Insufficiently stringent thresholds may fail to distinguish meaningful clinical change from natural fluctuation in disease activity.

  • outcomes research
  • disease activity
  • treatment
http://creativecommons.org/licenses/by-nc/4.0/

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.

https://doi.org/10.1136/lupus-2019-000364

Statistics from Altmetric.com

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

    Introduction

    Cutaneous lupus erythematosus (CLE) is an autoimmune skin disorder, which can occur in the context of SLE or independent of other organ involvement.1–3 Its clinical manifestations, severity and course are highly variable. This variability confounds the development of appropriate outcome measures that are reproducible, reflect the range of patient experience and reliably distinguish meaningful clinical improvement from fluctuation intrinsic to the natural history of the disease. As a result, studies have differed on defining clinical improvement in CLE. Outcomes have been measured using subjective assessments of improvement4 5 and different semiquantitative severity scoring systems.6–8 In the absence of clear outcome measures, assessing the effectiveness of different therapies and selecting the most appropriate treatments for individual patients has been challenging. While a variety of treatment options are available for CLE, treatment selection remains largely based on expert opinion rather than objective data.

    The most commonly used scoring system for CLE is the Cutaneous Lupus Activity and Severity Index (CLASI), which independently grades manifestations of CLE disease activity (CLASI-A), such as erythema and scaling, and skin damage (CLASI-D), such as dyspigmentation and scarring.9–11 In validation studies, CLASI demonstrates high inter-rater and intrarater reliability and correlates well with subjective physician and patient global assessments of disease burden.9 12 However, there is little consensus on how changes in CLASI scores should be used to classify treatment response. Prior studies have found four-point or 20% decrease in CLASI-A score on a 70-point scale to be indicative of noticeable clinical improvement.13 Despite the possibility that such modest changes may be less meaningful for patients with more severe involvement or may fail to distinguish treatment response from expected clinical variability, similar thresholds have been used to classify activity improvement in interventional and observational studies.8 14 Other endpoints used include larger relative changes in CLASI-A scores (eg, 50% improvement in CLASI-A),15 16 analogous to the Psoriasis Area Severity Index (PASI) percentage change endpoints common in psoriasis studies (eg, PASI50).17 Even less information is available regarding CLASI-D endpoints, as skin damage phenomena are generally regarded as permanent. However, modest improvement in CLASI-D scores has been observed in prior studies.9 18 19Because individual studies tend to rely on single CLASI thresholds to define clinical improvement, the impact of the particular threshold selected remains unclear. Just as a diagnostic test’s cut-off value affects the test’s sensitivity and specificity, the outcome threshold used with a disease severity scoring system will affect the performance of that scoring system in both observational and interventional studies. This effect has been observed in a number of other fields, including using body mass index thresholds to define obesity,20 blood pressure thresholds to define hypertension21 and serological testing thresholds to define chronic atrophic gastritis.22 Thus, defining how different CLASI thresholds influence models of CLE improvement is critically important for CLE study design.

    This study addresses that gap. Using longitudinal data from a cohort of patients enrolled in the University of Texas Southwestern (UTSW) Cutaneous Lupus Registry, we analysed CLE activity and damage improvement defined across ranges of relative change in CLASI-A and CLASI-D scores. By examining a range of outcome definitions rather than focusing on a single threshold to classify treatment response, we identified patient demographic and clinical factors associated with likelihood of CLE improvement across different thresholds. We found that the strengths of these associations depend strongly on the CLASI improvement thresholds used. In particular, we found that less stringent thresholds may not reliably distinguish meaningful clinical improvement from expected fluctuations in disease course, particularly for CLASI-A. We also identified association patterns that will aid providers in setting expectations when counselling patients. These findings have important implications for our understanding of CLE patient response to treatment and help guide the design of future clinical trials.

    Methods

    The UTSW Cutaneous Lupus Registry is a longitudinal cohort study of patients with CLE recruited from the outpatient dermatology clinics of UTSW and Parkland Health and Hospital System. This study was approved by the UTSW Institutional Review Board. Patients or the public were not involved in the design, or conduct, or reporting, or dissemination of this research. Data collected include patient demographics, disease history, medication history, laboratory data and CLE disease severity using CLASI scores. For primary outcome measures of skin disease activity and damage improvement, we designated response thresholds based on percent change in CLASI-A and CLASI-D scores between two consecutive study visits, defined as a visit-pair. Specifically,

    Embedded Image

    Embedded Image

    This approach was intended to target periods of approximately 6–12 months, a typical time course for evaluating treatment response or failure in practice and a common duration for clinical trials.7 23–25 Two response thresholds were initially selected to screen factors for each analysis. These included standard thresholds, representing relative decreases in CLASI-A (≥50%) or CLASI-D (≥20%) within the range of outcomes considered in prior studies,8 14–16 19 and stringent thresholds, representing larger relative decreases in CLASI-A (≥75%) or CLASI-D (≥40%) score.

    Patients with at least two study visits between July 2009 and September 2016 were available for inclusion in this study. Patients with drug-induced CLE were excluded. We designated two study subsets from this cohort. Patients with CLASI-A ≥5 or CLASI-D ≥5 at the initial visit of a visit-pair were selected for analysis of CLE disease activity or damage improvement, respectively, thus excluding patients with disease activity or skin damage levels too low to detect noticeable improvement.

    Statistical analysis was conducted using R software (version 3.4.4; https://www.R-project.org/) with the geepack package (version 1.2.1).26 Single-variable and multivariable logistic regression models were fit using generalised estimating equations with independence working correlation structures and robust variance estimators to account for potential lack of independence among visit-pairs from the same patient.27 As a sensitivity analysis, final multivariable models were also fit using exchangeable and first-order autoregressive correlation structures, which did not substantially impact the results. Single-variable models were used to screen potential independent variables selected based on literature review and clinical experience (online supplementary table 1). Factors with p<0.10 at either of the screening thresholds were included in the multivariable models, which were then fit across the response threshold ranges. Variation in parameter coefficients was characterised using Pearson product moment correlation, with the squared coefficient (R 2) representing the fraction of variation in the multivariable model parameter coefficient explained by linear change in the response threshold. P<0.05 was considered statistically significant.

    Supplemental material

    Results

    Study population

    Data from 66 patients with 119 consecutive visit-pairs were included in the analysis of CLASI-A improvement, and data from 74 patients with 177 consecutive visit-pairs were included in the analysis of CLASI-D improvement (table 1). The two analysis groups partially overlapped; 52 patients with a total of 89 visit-pairs were included in both analyses. Both groups were predominantly women with median ages of 46 years for the CLASI-A analysis and 45 years for the CLASI-D analysis. More than half of the patients in each group met diagnostic criteria for SLE. However, compared with the CLASI-A analysis, the CLASI-D analysis included a higher proportion of patients who were African-American (62% vs 49%) and had predominantly chronic CLE (CCLE) lesions (100% vs 85%). The range in disease activity change from visit to visit was quite broad; 35 (29%) visit-pairs had a relative CLASI-A decrease of ≥50%, and 17 (14%) had a relative CLASI-A decrease of ≥75% (figure 1). In contrast, damage changes were smaller; 37 (21%) visit-pairs had a relative CLASI-D decrease of ≥20%, and 18 (10%) had a relative CLASI-D decrease of ≥40% (figure 1). Several factors demonstrated potential positive or negative associations with CLASI-A and CLASI-D improvement based on cross-tabulations and single-variable models (online supplementary tables 1 and 2 and figure 1). Factors with p<0.10 in single-variable screening models, including smoking, CLE subtype, and age of CLE development for CLASI-A improvement and initial CLASI-A score, African-American race, and positive ANA for CLASI-D improvement, were selected for inclusion in the multivariable models.

    Supplemental material

    Figure 1
    Figure 1

    Percent change in Cutaneous Lupus Activity and Severity Index (CLASI) scores. Distribution of per cent change in CLASI activity (CLASI-A, top) and CLASI damage (CLASI-D, bottom) scores within visit-pairs. Blue line delineates ≥50% decrease in CLASI-A and ≥20% decrease in CLASI-D. Red line delineates ≥75% decrease in CLASI-A and ≥40% decrease in CLASI-D.

    View this table:
    Table 1

    Demographic and baseline characteristics

    Analysis across response thresholds

    To more completely characterise the impact of outcome measure selection on factors associated with activity and damage improvement, we compared multivariable logistic regression models across ranges of CLASI-A and CLASI-D response thresholds. The impact of these thresholds was substantial, as demonstrated by differences between models at the predefined screening thresholds of ≥50% CLASI-A improvement, ≥75% CLASI-A improvement, ≥20% CLASI-D improvement and ≥40% CLASI-D improvement (table 2). In particular, subacute cutaneous lupus erythematosus (SCLE) predominance was associated with increased likelihood of CLASI-A improvement only at the 75% threshold (OR 5.67 (95% CI 1.56 to 20.5), p=0.008). Additionally, African-American race was modestly associated with decreased likelihood of CLASI-D improvement at the 20% threshold (OR 0.40 (95% CI 0.17 to 0.93), p=0.03) but more significantly at the 40% threshold (OR 0.25 (95% CI 0.08 to 0.82), p=0.02). In contrast, age of CLE development was similarly associated with a higher likelihood of CLASI-A improvement at the 50% threshold (OR per 10 year increase: 1.612 (95% CI 1.127 to 2.304), p=0.009) and at the 75% threshold (OR 1.616 (95% CI 1.015 to 2.574), p=0.04). Initial CLASI-A score was similarly associated with lower likelihood of CLASI-D improvement at the 20% threshold (OR per 5-point increase: 0.625 (95% CI 0.427 to 0.915), p=0.02) and at the 40% threshold (OR 0.514 (95% C I0.312 to 0.846), p=0.009).

    View this table:
    Table 2

    Multivariable models

    Model fits across the full CLASI-A and CLASI-D response threshold ranges (figure 2 and online supplementary table 3) further illustrate the considerable variation in strength of association between different independent variables and the likelihood of CLE activity or damage improvement, depending on the particular CLASI-A or CLASI-D threshold used. In figure 2, ORs for each parameter in the multivariable models are individually compared across thresholds. For some factors, the strength of association with likelihood of activity or damage improvement increased or decreased with greater relative changes in CLASI-A and CLASI-D. This was the case for SCLE predominance (R 2=0.73) in the CLASI-A analysis, where the strength of association increased with greater improvement in CLASI-A scores. In African-American patients (R 2=0.80) and those with history of positive ANA (R 2=0.91) in the CLASI-D analysis, the strength of association with damage improvement decreased with greater relative changes in CLASI-D, although the latter association was quite weak. In contrast, other factors showed more stable associations throughout the spectrum of relative changes of CLASI-A and CLASI-D scores. These included current smoking history (R 2=0.30) and age CLE developed (R 2=0.25) in the CLASI-A analysis and initial disease activity (R 2=0.55) in the CLASI-D analysis.

    Figure 2
    Figure 2

    Associations between patient factors and cutaneous lupus erythematosus activity and damage improvement depend on the response threshold. Each panel compares ORs for a single independent variable in multivariable logistic regression models across different thresholds for activity or damage improvement, defined as per cent change in Cutaneous Lupus Activity and Severity Index (CLASI) Activity (CLASI-A, top row) or Damage (CLASI-D, bottom row). ORs (points) with 95% CI (vertical lines) are plotted on the y-axis, with the position on the x-axis corresponding to the CLASI-A or CLASI-D per cent change threshold. Red curves indicate linear OR trends. CLE, chronic cutaneous lupus erythematosus; SCLE, subacute cutaneous lupus erythematosus.

    Exploratory secondary analyses

    In order to fully explore potential associations in the dataset, we fit expanded multivariable models for CLASI-A and CLASI-D improvement including factors with p<0.20 in the single-variable models (online supplementary table 4). The expanded models identified several potential associations of interest, including use of immunosuppressant medications at baseline and decreased likelihood of CLASI-A improvement at the 50% (OR 0.246 (95% CI 0.083 to 0.732)) and 75% thresholds (OR 0.112 (95% CI 0.022 to 0.571)), and longer follow-up interval with increased likelihood of ≥40% CLASI-D improvement (OR per month: 1.042 (95% CI 1.007 to 1.078)). However, given the large number of parameters included in these models relative to the number of events, more data will be needed to confirm these associations.28 29

    As SCLE predominance had the strongest association with ≥75% CLASI-A improvement in the multivariable model (table 2), and the pathophysiology of CLE subtypes may differ, we separated patients in the CLASI-A analysis into subgroups based on those with predominantly SCLE or predominantly CCLE lesions (online supplementary table 5). Compared with patients with CCLE, patients with SCLE developed CLE at an older age, another factor associated with increased likelihood of CLASI-A improvement in the multivariable models. However, older age of CLE development remained associated with CLASI-A improvement in a single-variable model of the CCLE subgroup at the 50% (OR per 10 year increase: 1.57 (95% CI 1.07 to 2.30), p=0.02) and 75% thresholds (1.58 (95% CI 0.87 to 2.89), p=0.13), although the latter did not reach statistical significance due to small sample size. Since stratifying by CLE subtype does not completely explain the association between age of CLE development and CLASI-A improvement, this supports inclusion of both CLE subtypes in the multivariable models.

    Discussion

    The relationships between patient factors and CLE activity and skin damage improvement depend on the CLASI-A and CLASI-D reduction thresholds used to classify clinical improvement. For example, SCLE predominance for activity improvement and African-American race for damage improvement had little association at lower changes in CLASI-A and CLASI-D but developed much stronger associations at higher changes, particularly ≥75% compared with ≥50% CLASI-A improvement and ≥40% compared with ≥20% CLASI-D improvement. These findings highlight the importance of evaluating a range of CLASI improvement threshold outcome measures to obtain a comprehensive picture of disease prognosis. There may be many difficult-to-identify factors influencing small to moderate changes in CLASI score, including the natural waxing and waning course of CLE as well as CLASI measurement uncertainty. Insufficiently stringent CLASI thresholds may thus fail to distinguish meaningful clinical benefit from other causes of fluctuation in disease severity. Future studies of potential CLE treatments can evaluate multiple thresholds that include large improvements in disease activity (eg, ≥75% CLASI-A improvement) rather than focusing solely on one threshold (eg, ≥50% CLASI-A improvement) and be powered accordingly.

    Varying effects across threshold values can also reveal important properties of the underlying associations and may suggest separate processes influencing moderate compared with significant CLE improvement. For example, SCLE predominance was associated with CLASI-A improvement only at stringent thresholds, suggesting that while patients with SCLE or CCLE lesions may experience moderate activity improvement, dramatic short-term improvement is more common for SCLE than for CCLE. One possibility is that this divergence may reflect the relatively sparse superficial perivascular lymphocytic infiltrate in SCLE, compared with the denser perifollicular and perivascular inflammatory infiltrate in CCLE, particularly in discoid lesions.30 This information can be useful to providers, who can counsel patients with SCLE that they are more likely to experience significant improvement in their lesions with appropriate treatment. Additionally, African-American race was more strongly associated with decreased likelihood of ≥40% CLASI-D improvement compared with ≥20% CLASI-D improvement. This finding may reflect the potentially greater visibility of damage phenomena, particularly dyspigmentation, in darker skin types. This finding will also help providers set appropriate expectations for these patients regarding their disease damage course. Consistent with this result, other studies have demonstrated higher levels of CLE damage in African-American patients, although a higher prevalence of CCLE in African-Americans could confound this relationship.31 32

    Other predictor variables showed varying patterns across different CLASI-A or CLASI-D thresholds. Age of CLE development was modestly associated with increased likelihood of activity improvement at both the ≥50% and ≥75% CLASI-A thresholds, although increasing uncertainty at more stringent thresholds complicates the interpretation. This result suggests that patients who develop CLE early in life may be less responsive to treatment than those with later-onset disease. In a similar pattern, other studies have found that early-onset SLE patients present with more activity than those with late-onset disease due to higher frequency of major organ involvement33 34 and may also be less responsive to treatment.35 However, while younger age of CLE development was associated with decreased likelihood of CLE activity improvement between visits in our analyses, it was not associated with higher CLASI-A scores (data not shown). Additionally, initial CLE activity was similarly associated with lower likelihood of CLE damage improvement at both the ≥20% and ≥40% CLASI-D thresholds. This pattern fits with our understanding that CLE disease activity can later develop into manifestations of skin damage. Thus, patients with higher initial CLASI-A scores are less likely to experience even modest improvement in CLASI-D scores between visits.

    While the varying effect sizes across CLASI improvement thresholds found in this study demonstrate the importance of examining a range of thresholds, they also present two potential pitfalls. First, varying effect sizes across thresholds should not be exploited to select one threshold, which maximises the effect of a particular parameter of interest. Such optimisation methods will significantly bias results of the analysis to inflate the effect size of the selected parameter. Additionally, when relying on a null hypothesis significance testing framework, care should be taken to manage potential type I error rate inflation from multiple comparisons. Appropriate strategies could include prespecification of a small number of primary outcome thresholds, as we have done in this study, or statistical methods for controlling global level type I error rates, such as the Bonferroni correction.36

    Limitations of this study include small sample size and selection bias due to the study occurring at a single tertiary centre. In addition, while many study participants started or stopped medications, these changes were distributed over a wide range of treatments. As a result, individual therapies could not be included in the analysis, and treatment changes were instead grouped by medication class (table 1). However, we did not find that adding or stopping classes of medications were significant factors in changes in CLASI activity or damage. Our study was also limited to analysing changes between visits, thus providing a relatively short-term perspective. One recent study has described longer term trends in patients with CLE but did not examine relative changes in CLASI-A or CLASI-D scores.19 Finally, this observational study cannot definitively determine the underlying causes of the identified associations and their variation with CLASI improvement thresholds, and future studies will be needed to better define these mechanisms.

    The factors associated with CLE activity and damage improvement identified in this study can inform patient counselling and treatment planning. For example, our data support the idea that patients with SCLE can reasonably expect significant improvements in their skin disease activity. Additionally, the variability of these associations depending on the CLASI reduction thresholds used to define clinical response highlights the need to examine a range of response thresholds in future studies. Individual threshold-based outcome measures may miss important effects or fail to adequately capture the clinical context. In particular, smaller changes in CLASI may occur in patients, but they may be more difficult to attribute to a particular predictor or intervention. Based on these results, larger multicentre studies currently being pursued can include evidence-based outcome measures that adequately assess therapeutic efficacy for CLE.

    Acknowledgments

    The authors would like to thank Dr Yulun Liu for reviewing the biostatistical analyses of our manuscript, and Drs Elaine Kunzler, Rebecca Vasquez, Andrew Kim, Daniel Grabell, Noelle Teske, Tina Vinoya, Jack O’Brien, Danielle Lin, Jenny Raman and Justin Raman for recruiting patients, and the participants of the University of Texas Southwestern (UTSW) CLE Registry for their contributions to lupus research.

    References

    1. ↵
      2. Gilliam JN ,
      3. Sontheimer RD
      . Distinctive cutaneous subsets in the spectrum of lupus erythematosus. J Am Acad Dermatol 1981;4:471–5.doi:10.1016/S0190-9622(81)80261-7
      OpenUrlCrossRefPubMedWeb of Science
    2. ↵
      2. Rothfield N ,
      3. Sontheimer RD ,
      4. Bernstein M
      . Lupus erythematosus: systemic and cutaneous manifestations. Clin Dermatol 2006;24:348–62.doi:10.1016/j.clindermatol.2006.07.014
      OpenUrlCrossRefPubMedWeb of Science
    3. ↵
      2. Tsokos GC
      . Systemic lupus erythematosus. N Engl J Med 2011;365:2110–21.doi:10.1056/NEJMra1100359
      OpenUrlCrossRefPubMedWeb of Science
    4. ↵
      2. Wahie S ,
      3. Daly AK ,
      4. Cordell HJ , et al
      . Clinical and pharmacogenetic influences on response to hydroxychloroquine in discoid lupus erythematosus: a retrospective cohort study. J Invest Dermatol 2011;131:1981–6.doi:10.1038/jid.2011.167
      OpenUrlCrossRefPubMedWeb of Science
    5. ↵
      2. Gammon B ,
      3. Hansen C ,
      4. Costner MI
      . Efficacy of mycophenolate mofetil in antimalarial-resistant cutaneous lupus erythematosus. J Am Acad Dermatol 2011;65:717–21.doi:10.1016/j.jaad.2010.08.011
      OpenUrlCrossRefPubMed
    6. ↵
      2. Barikbin B ,
      3. Givrad S ,
      4. Yousefi M , et al
      . Pimecrolimus 1% cream versus betamethasone 17-valerate 0.1% cream in the treatment of facial discoid lupus erythematosus: a double-blind, randomized pilot study. Clin Exp Dermatol 2009;34:776–80.doi:10.1111/j.1365-2230.2008.03138.x
      OpenUrlCrossRefPubMed
    7. ↵
      2. Kuhn A ,
      3. Landmann A ,
      4. Patsinakidis N , et al
      . Fumaric acid ester treatment in cutaneous lupus erythematosus (CLE): a prospective, open-label, phase II pilot study. Lupus 2016;25:1357–64.doi:10.1177/0961203316644335
      OpenUrlCrossRefPubMed
    8. ↵
      2. Khamashta M ,
      3. Merrill JT ,
      4. Werth VP , et al
      . Sifalimumab, an anti-interferon-α monoclonal antibody, in moderate to severe systemic lupus erythematosus: a randomised, double-blind, placebo-controlled study. Ann Rheum Dis 2016;75:1909–16.doi:10.1136/annrheumdis-2015-208562
      OpenUrlAbstract/FREE Full Text
    9. ↵
      2. Bonilla-Martinez ZL ,
      3. Albrecht J ,
      4. Troxel AB , et al
      . The cutaneous lupus erythematosus disease area and severity index: a responsive instrument to measure activity and damage in patients with cutaneous lupus erythematosus. Arch Dermatol 2008;144:173–80.doi:10.1001/archderm.144.2.173
      OpenUrlCrossRefPubMedWeb of Science
    10. ↵
      2. Klein RS ,
      3. Morganroth PA ,
      4. Werth VP
      . Cutaneous lupus and the cutaneous lupus erythematosus disease area and severity index instrument. Rheum Dis Clin North Am 2010;36:33–51.doi:10.1016/j.rdc.2009.12.001
      OpenUrlPubMed
    11. ↵
      2. Albrecht J ,
      3. Taylor L ,
      4. Berlin JA , et al
      . The CLASI (cutaneous lupus erythematosus disease area and severity index): an outcome instrument for cutaneous lupus erythematosus. J Invest Dermatol 2005;125:889–94.doi:10.1111/j.0022-202X.2005.23889.x
      OpenUrlCrossRefPubMedWeb of Science
    12. ↵
      2. Jolly M ,
      3. Kazmi N ,
      4. Mikolaitis RA , et al
      . Validation of the cutaneous lupus disease area and severity index (CLASI) using physician- and patient-assessed health outcome measures. J Am Acad Dermatol 2013;68:618–23.doi:10.1016/j.jaad.2012.08.035
      OpenUrlCrossRefPubMed
    13. ↵
      2. Klein R ,
      3. Moghadam-Kia S ,
      4. LoMonico J , et al
      . Development of the CLASI as a tool to measure disease severity and responsiveness to therapy in cutaneous lupus erythematosus. Arch Dermatol 2011;147:203–8.doi:10.1001/archdermatol.2010.435
      OpenUrlCrossRefPubMedWeb of Science
    14. ↵
      2. Parodis I ,
      3. Gomez A ,
      4. Frodlund M , et al
      . Smoking reduces the efficacy of belimumab in mucocutaneous lupus. Expert Opin Biol Ther 2018;18:911–20.doi:10.1080/14712598.2018.1494719
      OpenUrl
    15. ↵
      2. Furie R ,
      3. Khamashta M ,
      4. Merrill JT , et al
      . Anifrolumab, an Anti-Interferon-α receptor monoclonal antibody, in moderate-to-severe systemic lupus erythematosus. Arthritis Rheumatol 2017;69:376–86.doi:10.1002/art.39962
      OpenUrl
    16. ↵
      2. van Vollenhoven RF ,
      3. Hahn BH ,
      4. Tsokos GC , et al
      . Efficacy and safety of ustekinumab, an IL-12 and IL-23 inhibitor, in patients with active systemic lupus erythematosus: results of a multicentre, double-blind, phase 2, randomised, controlled study. Lancet 2018;392:1330–9.doi:10.1016/S0140-6736(18)32167-6
      OpenUrl
    17. ↵
      2. Weisman S ,
      3. Pollack CR ,
      4. Gottschalk RW
      . Psoriasis disease severity measures: comparing efficacy of treatments for severe psoriasis. J Dermatolog Treat 2003;14:158–65.doi:10.1080/09546630310013360
      OpenUrlCrossRefPubMed
    18. ↵
      2. Erceg A ,
      3. Bovenschen HJ ,
      4. van de Kerkhof PCM , et al
      . Efficacy and safety of pulsed dye laser treatment for cutaneous discoid lupus erythematosus. J Am Acad Dermatol 2009;60:626–32.doi:10.1016/j.jaad.2008.11.904
      OpenUrlCrossRefPubMedWeb of Science
    19. ↵
      2. Ker KJ ,
      3. Teske NM ,
      4. Feng R , et al
      . Natural history of disease activity and damage in patients with cutaneous lupus erythematosus. J Am Acad Dermatol 2018;79:1053–60.doi:10.1016/j.jaad.2018.06.040
      OpenUrl
    20. ↵
      2. Breitling LP ,
      3. Brenner H
      . Odd odds interactions introduced through dichotomisation of continuous outcomes. J Epidemiol Community Health 2010;64:300–3.doi:10.1136/jech.2009.089458
      OpenUrlAbstract/FREE Full Text
    21. ↵
      2. Ragland DR
      . Dichotomizing continuous outcome variables: dependence of the magnitude of association and statistical power on the cutpoint. Epidemiology 1992;3:434–40.doi:10.1097/00001648-199209000-00009
      OpenUrlCrossRefPubMedWeb of Science
    22. ↵
      2. Brenner H ,
      3. Rothenbacher D ,
      4. Weck MN
      . Epidemiologic findings on serologically defined chronic atrophic gastritis strongly depend on the choice of the cutoff-value. Int J Cancer 2007;121:2782–6.doi:10.1002/ijc.22992
      OpenUrlCrossRefPubMedWeb of Science
    23. ↵
      2. Werth VP ,
      3. Fiorentino D ,
      4. Sullivan BA , et al
      . Brief report: pharmacodynamics, safety, and clinical efficacy of AMG 811, a human Anti-Interferon-γ antibody, in patients with discoid lupus erythematosus. Arthritis Rheumatol 2017;69:1028–34.doi:10.1002/art.40052
      OpenUrl
    24. ↵
      2. Okon L ,
      3. Rosenbach M ,
      4. Krathen M , et al
      . Lenalidomide in treatment-refractory cutaneous lupus erythematosus: efficacy and safety in a 52-week trial. J Am Acad Dermatol 2014;70:583–4.doi:10.1016/j.jaad.2013.11.007
      OpenUrlCrossRefPubMed
    25. ↵
      2. Islam MN ,
      3. Hossain M ,
      4. Haq SA , et al
      . Efficacy and safety of methotrexate in articular and cutaneous manifestations of systemic lupus erythematosus. Int J Rheum Dis 2012;15:62–8.doi:10.1111/j.1756-185X.2011.01665.x
      OpenUrlCrossRefPubMed
    26. ↵
      2. Halekoh U ,
      3. Højsgaard S ,
      4. Yan J
      . The R Package geepack for Generalized Estimating Equations. J Stat Softw 2006;15.doi:10.18637/jss.v015.i02
    27. ↵
      2. Yan J ,
      3. Fine J
      . Estimating equations for association structures. Stat Med 2004;23:859–74. discussion 75-7,79-80.doi:10.1002/sim.1650
      OpenUrlCrossRefPubMedWeb of Science
    28. ↵
      2. Pan W
      . Akaike's information criterion in generalized estimating equations. Biometrics 2001;57:120–5.doi:10.1111/j.0006-341X.2001.00120.x
      OpenUrlCrossRefPubMedWeb of Science
    29. ↵
      2. Hin L-Y ,
      3. Wang Y-G
      . Working-correlation-structure identification in generalized estimating equations. Stat Med 2009;28:642–58.doi:10.1002/sim.3489
      OpenUrlCrossRefPubMedWeb of Science
    30. ↵
      2. Baltaci M ,
      3. Fritsch P
      . Histologic features of cutaneous lupus erythematosus. Autoimmun Rev 2009;8:467–73.doi:10.1016/j.autrev.2008.12.014
      OpenUrlCrossRefPubMed
    31. ↵
      2. Verma SM ,
      3. Okawa J ,
      4. Propert KJ , et al
      . The impact of skin damage due to cutaneous lupus on quality of life. Br J Dermatol 2014;170:315–21.doi:10.1111/bjd.12653
      OpenUrl
    32. ↵
      2. Pons-Estel GJ ,
      3. Alarcón GS ,
      4. González LA , et al
      . Possible protective effect of hydroxychloroquine on delaying the occurrence of integument damage in lupus: LXXI, data from a multiethnic cohort. Arthritis Care Res (Hoboken) 2010;62:393–400.doi:10.1002/acr.20097
      OpenUrl
    33. ↵
      2. Livingston B ,
      3. Bonner A ,
      4. Pope J
      . Differences in autoantibody profiles and disease activity and damage scores between childhood- and adult-onset systemic lupus erythematosus: a meta-analysis. Semin Arthritis Rheum 2012;42:271–80.doi:10.1016/j.semarthrit.2012.05.001
      OpenUrlCrossRefPubMed
    34. ↵
      2. Livingston B ,
      3. Bonner A ,
      4. Pope J
      . Differences in clinical manifestations between childhood-onset lupus and adult-onset lupus: a meta-analysis. Lupus 2011;20:1345–55.doi:10.1177/0961203311416694
      OpenUrlCrossRefPubMedWeb of Science
    35. ↵
      2. Feng X ,
      3. Zou Y ,
      4. Pan W , et al
      . Associations of clinical features and prognosis with age at disease onset in patients with systemic lupus erythematosus. Lupus 2014;23:327–34.doi:10.1177/0961203313513508
      OpenUrlCrossRefPubMedWeb of Science
    36. ↵
      2. Bender R ,
      3. Lange S
      . Adjusting for multiple testing--when and how? J Clin Epidemiol 2001;54:343–9.doi:10.1016/S0895-4356(00)00314-0
      OpenUrlCrossRefPubMedWeb of Science

    Footnotes

    • Contributors BAN and BFC designed the study and drafted the manuscript. All authors were responsible for acquisition, analysis and interpretation of the data. BAN performed the statistical analysis. BFC obtained funding and supervised the study.

    • Funding This study was supported in part by the National Institute of Arthritis and Musculoskeletal and Skin Diseases of the National Institutes of Health under Award Number K23AR061441.

    • Disclaimer The content is solely the responsibility of the authors and does not necessarily represent the official views of the University of Texas Southwestern Medical Center at Dallas and its affiliated academic and health care centers or the National Institutes of Health.

    • Competing interests BFC is an investigator for Daavlin Corporated, Biogen Incoporated and Pfizer Incorporated. He is a consultant for Viela Bio.

    • Patient consent for publication Not required.

    • Ethics approval This study was approved by the UTSW Institutional Review Board.

    • Provenance and peer review Not commissioned; externally peer reviewed.

    • Data availability statement Data are available on reasonable request. All data relevant to the study are included in the article or uploaded as supplementary information.