Measuring Disease Damage and Its Severity in Childhood-Onset Systemic Lupus Erythematosus

Abstract

Objective.

To describe the frequency and types of disease damage occurring with childhood-onset systemic lupus erythematosus (SLE) as measured by the 41-item Systemic Lupus International Collaborating Clinics/American College of Rheumatology Damage Index (SDI), and to assess the SDI’s ability to reflect damage severity.

Methods.

Information for the SDI was prospectively collected from 1,048 childhood-onset SLE patients. For a subset of 559 patients, physician-rated damage severity measured by visual analog scale (MD VAS damage) was also available. Frequency of SDI items and the association between SDI summary scores and MD VAS damage were estimated. Finally, an international consensus conference, using nominal group technique, considered the SDI’s capture of childhood-onset SLE–associated damage and its severity.

Results.

After a mean disease duration of 3.8 years, 44.2% of patients (463 of 1,048) already had an SDI summary score >0 (maximum 14). The most common SDI items scored were proteinuria, scarring alopecia, and cognitive impairment. Although there was a moderately strong association between SDI summary scores and MD VAS damage (Spearman’s r = 0.49, P < 0.0001) in patients with damage (SDI summary score >0), mixed-effects analysis showed that only 4 SDI items, each occurring in <2% of patients overall, were significantly associated with MD VAS damage. There was consensus among childhood-onset SLE experts that the SDI in its current form is inadequate for estimating the severity of childhood-onset SLE–associated damage.

Conclusion.

Disease damage as measured by the SDI is common in childhood-onset SLE, even with relatively short disease durations. Given the shortcomings of the SDI, there is a need to develop new tools to estimate the impact of childhood-onset SLE–associated damage.

INTRODUCTION

Disease activity describes theoretically reversible manifestations due to the inflammatory processes underlying systemic lupus erythematosus (SLE), while the term damage is used to designate irreversible organ scarring or tissue degradation. However, a formal consensus definition of disease damage with SLE has not been published. Quantifying damage or measuring the severity of damage is an important consideration in gauging the overall outcome in SLE, particularly as disease-related mortality decreases (1,2). Given the diversity of SLE-associated organ involvement, and in line with the development of disease activity indices for quantifying disease activity, an SLE damage index (the Systemic Lupus International Collaborating Clinics/American College of Rheumatology Damage Index [SDI]) has been developed (1).

Prior investigations support the idea that the presence of SLE-associated damage that is scored by the SDI is associated with increased mortality in adults (2) and with increased cumulative SLE activity in both adults and children (2–4). The developers of the SDI stressed that the index provides an enumeration of the presence or absence of damage types only, and that it does not directly quantify damage severity (3). This distinction is sensible because, for example, a patient with a small cataract and another with a debilitating stroke will both receive the same SDI summary score of 1, provided each patient fulfills the definition of only 1 SDI item. Nevertheless, the SDI summary score has been used as an independent, continuous outcome measure in statistical analyses (5).

There have been prior attempts to adapt the SDI to better measure damage severity by introducing item weightings. However, 2 prior attempts at item weightings using data gathered from adults with SLE did not meaningfully improve the association of SDI summary scores with mortality, and therefore item weightings were not pursued further (2,6). One prior attempt was made to weight SDI items (based on the Systemic Lupus Erythematosus Disease Activity Index item weights) in data from patients with childhood-onset SLE; this approach did not improve prediction of damage using cumulative disease activity as a predictor and was similarly not pursued (4).

In approximately 20% of patients with SLE, the disease is diagnosed during childhood (childhood-onset SLE), i.e., ages <18 years (7). A prior international consensus process (8), focused on outcome measures in childhood-onset SLE and juvenile dermatomyositis, defined a core set of variables to characterize childhood-onset SLE–associated damage and its impact. Based on consensus, the agreed-upon childhood-onset SLE damage core set consists of the SDI as the current standard damage tool, a physician global assessment of damage (visual analog scale [VAS] or Likert scale), growth (height and weight), bodily development (menses, Tanner staging), and a measure of health-related quality of life.

The Outcome Measures in Rheumatology (OMERACT) collaborative has recently published a revised framework. The OMERACT filter 2.0 builds on the OMERACT filter 1.0, i.e., the use of rheumatology outcome measures that are valid, discriminate conditions of interest, and are feasible (9). The OMERACT filter 2.0 advocates development of core outcome measurement sets in rheumatology (10). There is new emphasis on the concept of multiple core areas (death, disease impact, resource use, and disease manifestations), which may be addressed by these core outcome sets. Finally, the need to explicitly consider perspective is stressed, as well as the context in which an outcome measure will be used. The aspect of perspective (patient, physician, and/or society) and context appear particularly important when capturing damage severity. While the previously developed childhood-onset SLE damage core set (8) touches on at least 2 of the OMERACT core areas for construct measurement, stand-alone use of the SDI is still commonplace when quantifying the amount of damage in clinical studies of both SLE and childhood-onset SLE.

Based on the above, we sought to critically appraise the SDI when used in childhood-onset SLE by delineating the frequency of SDI damage items in a large composite cohort under consideration of disease duration, by comparing the SDI summary score to a physician global assessment of damage severity, and by exploring the possible impact of SDI item weightings to better capture damage severity as rated by the treating physician using statistical techniques. Subsequently, our findings and analyses were presented at a consensus conference held in April 2017, where an international panel of childhood-onset SLE experts was asked to advise on approaches to measure damage severity in childhood-onset SLE.

MATERIALS AND METHODS

Patients.

Longitudinal data from large, prospective cohorts of childhood-onset SLE were reanalyzed. These were the United Kingdom Juvenile-onset SLE Cohort Study (UK; n = 350) (1), the childhood-onset SLE cohort followed at Cincinnati Children’s Hospital Medical Center (CCHMC; n = 139), and an international cohort assembled in Latin America, Australia, Asia, and various European countries by the Paediatric Rheumatology International Trials Organisation (PRINTO; n = 559) (8). The final composite study cohort included 1,048 patients. General demographic data were recorded, though ethnic/racial data were not collected for the PRINTO data set due to legal restrictions. Approval was given by the CCHMC Institutional Review Board for this secondary analysis. Training had been provided for the completion of the SDI and other disease measures of all cohorts considered in this study. All physicians completing instruments were experienced in the care of children with childhood-onset SLE.

Damage measures and scales, SDI.

Developed through a consensus process focused on adults with SLE, the SDI quantifies irreversible damage according to specific item definitions (1). The SDI captures damage in patients since their diagnosis with SLE, irrespective of whether damage is due to the SLE process itself, its treatment, or a comorbid condition. Damage captured in the SDI is considered nonreversible if any given item has been present for at least 6 months continuously, or immediately for some events associated with acute organ damage, e.g., myocardial infarct (1).

The 41 items included in the SDI were selected based on experience with adult-onset SLE, with items mostly scored as being either present or absent. While there is no overall weighting system, 6 SDI items (stroke, myocardial infarct, tissue loss, bowel infarct, avascular necrosis, or malignancy) can be scored twice, if 2 qualifying events occur at least 6 months apart from each other. The SDI item end stage renal disease is always given a score of 3 when present for at least 6 months continuously. Scoring is cumulative, and once an item qualifies for scoring in the SDI, that item is always scored moving forward, even if it subsequently resolves or is corrected (2). The SDI summary score is the simple sum of the item scores; thus, a summary score of 0 is assigned to patients who have never met criteria for any listed damage item.

VAS of disease damage severity.

Treating physicians contributing data to the PRINTO cohort rated damage severity on a 10-cm VAS (MD VAS damage) for their patients. The VAS used a 10-cm scale (where 0 = no damage and 10 = very severe damage), and the rating was done either prior to or after completing the SDI.

Statistical analyses.

Descriptive analyses included frequencies for categorical variables as well as means ± SDs for numerical variables. Comparison of numerical demographic features and SDI summary scores between cohorts was done by a fixed-effects model, with differences of post hoc means between groups corrected for multiple comparisons using Tukey’s method. Damage item frequency was calculated based on the SDI score at the last followup visit available for each patient, and per SDI instructions, included all SDI items ever scored for a given patient. Contingency table analysis compared item frequencies between groups of patients (UK, CCHMC, PRINTO). Significant differences between item frequencies were based on chi-square analyses or Fisher’s exact test, where appropriate. Given variation in disease duration, for purposes of statistical comparison of item frequency between cohorts, only patient SDI scores with total disease durations ≤4.5 years were considered. The 4.5-year cutoff reflects the 70th percentile of followup in the cohort with the shortest followup (PRINTO).

To assess the relationship of SDI summary scores to physician-perceived severity of damage (MD VAS damage), we calculated Spearman’s correlation coefficients from patient visits with available data. Correlation with MD VAS damage was then separately assessed only for patient visits with known damage (SDI scores >0). This assessment was done to capture damage severity, which seems relevant only with the presence of some damage (SDI summary score >0). To evaluate the effect of multiple comparisons, correlations were also calculated separately using only the first and then last visit for each patient. Finally, individual damage items (present versus absent) were assessed for their associations with the MD VAS damage ratings, using logistic regression models and the generalized estimating equation method in computation. Computations were performed using SAS software, version 9.3, or Excel, version 2013.

Consensus conference.

An international consensus conference of physicians with expertise in childhood-onset SLE was held in Cincinnati, Ohio, from April 23–25, 2017. The expert group consisted of 13 physicians (10 pediatric rheumatologists and 3 pediatric nephrologists) with extensive experience in the care of childhood-onset SLE. Guided by an experienced moderator (BMF), nominal group technique was used to facilitate discussion and consensus formation. Consensus was defined a priori as ≥75% agreement among participating experts. Prior to opening discussion, results of the analyses (see below) and a review of relevant literature were presented.

RESULTS

Patient results.

As expected, most patients included in this study were female (82.9%), without significant sex differences between cohorts. Although UK and CCHMC cohort data were collected after 2006, the PRINTO cohort data were completed by 2004. There were significant racial differences present between the UK and CCHMC cohorts. Patient age at diagnosis and mean total disease duration were significantly higher in the CCHMC cohort when compared to the other data sets (CCHMC versus other cohorts, mean age 13.9 versus 12 years; P < 0.0001; mean duration 5.15 versus 3.5–3.8 years; P < 0.0001). Additional details are shown in Table 1.

Table 1.

Demographics and summary score by cohort^*

	All cohorts	UK	CCHMC	PRINTO	UK vs. CCHMC, P	UK vs. PRINTO, P	CCHMC vs. PRINTO, P
Total	1,048	350	139	559	–	NA	–
Female, no. (%)†	869 (82.9)	291 (83.1)	118 (84.9)	460 (82.3)	–	–	–
Age at diagnosis, years‡	12.2 ± 3.1	12 ± 3.32	13.9 ± 2.82	12 ± 2.95	< 0.0001	0.86	< 0.0001
Disease duration, years‡	3.81 ± 2.98	3.82 ± 3.17	5.15 ± 3.75	3.46 ± 2.57	< 0.0001	0.08	< 0.0001
SDI summary score‡	0.94 ± 1.56	0.56 ± 1.06	0.99 ± 1.59	1.16 ± 1.76	0.0006	0.0001	0.30
SDI summary score = 0, no. (%)†	585 (55.8)	272 (77.7)	80 (58.5)	233 (41.7)	–	< 0.0001	–

^*Values are the mean ± SD unless indicated otherwise.

UK = United Kingdom; CCHMC = Cincinnati Children’s Hospital Medical Center; PRINTO = Paediatric Rheumatology International Trials Organisation; NA = not applicable; SDI = Systemic Lupus International Collaborating Clinics/American College of Rheumatology Damage Index.

^†Values are the frequency; P values are from chi-square tests and only listed if P < 0.1.

^‡P values are from post-hoc analysis using fixed-effect models and correcting for the Tukey’s method, and are only listed if P < 0.1.

Overall disease damage as measured by the SDI.

Among the 1,048 patients, a total of 585 (55.8%) lacked disease damage (SDI summary score = 0) at the time of the last followup, which occurred, on average, 3.8 years post-diagnosis with childhood-onset SLE (Table 1). The proportion of patients without damage at last followup was highest in the UK cohort (77.7%). Mean SDI summary scores significantly differed between the UK and PRINTO cohorts, and PRINTO and CCHMC cohorts, but not between the CCHMC and UK cohorts (Table 1). Overall, the mean SDI summary score increased incrementally and was closely related to increasing disease duration (Figure 1A).

Figure 1.

Damage accumulation over time. A, The relation of the Systemic Lupus International Collaborating Clinics/American College of Rheumatology Damage Index (SDI) summary scores and disease duration in the cohort (n = 1,048) is shown for patients up to 10 years since the diagnosis with childhood-onset SLE. Analysis of variance supports a strong linear relationship (R² = 0.9, P < 0.0001). B, The relationship of the presence of damage in the 9 organ domains of the SDI is shown for disease durations up to 6 years. Diabetes mellitus and malignancies were very rare and are omitted. Lines are moving averages of annual values. The most commonly damaged organ systems were the neuropsychiatric, musculoskeletal, renal, and skin. Malignancies and diabetes mellitus were present in <0.4% of the composite cohort, and both SDI organ systems are excluded from the figure. ** = function of linear trend line, including information about model fit (R²); GI = gastrointestinal; MSK = musculoskeletal.

Common and less common types of disease damage with childhood-onset SLE.

The frequency of SDI items at the final followup visit in each cohort is shown in Table 2. The 3 most commonly encountered SDI items were long-standing nephrotic-range proteinuria, scarring alopecia, and chronic cognitive impairment in all 3 cohorts. Irrespective of disease duration, the 4 most commonly damaged organ systems were the neuropsychiatric, kidney, skin, and musculoskeletal (Figure 1B). Four SDI items were present in fewer than 3 patients (3 of 1,048 = 0.3%): angina, myocardial infarction, mesenteric insufficiency, and tendon rupture. An additional 12 SDI items were present in <1% of the study population. No patient had the SDI item pulmonary infarction scored (Table 2).

Table 2.

Comparison of SDI item frequency by cohort^*

Item	Overall	UK	CCHMC	PRINTO	P
Ocular
Cataracts	4.2	2	6.5	5	0.38
Retinal change	1.7	1.1	0.7	2.3	0.24
Neuropsychiatric
Cognitive impairment	8.4	3.4	12.2	10.7	0.0004†
Seizures requiring therapy	3.9	2.3	2.2	5.4	0.09†
Cerebrovascular accident	2.9/0.4	2 /−	3.6/1.4	3.4/0.4	0.014‡
Cranial/peripheral neuropathy	2.4	2.3	0.7	2.9	0.33‡
Transverse myelitis	0.4	0.3	0.7	0.4	0.29‡
Renal
Estimated GFR <50%	3.6	1.1	4.3	5	0.034†
Proteinuria >3.5 gm/day	9.6	4.3	10.8	12.7	0.0038†
End-stage renal disease	1.3	0.6	3.6	1.3	0.14‡
Pulmonary
Pulmonary hypertension	0.4	0.9	0.7	0	0.41‡
Pleural fibrosis	1.3	0.3	0	2.3	0.18‡
Shrinking lung	0.8	0.3	0	1.3	0.82‡
Pulmonary fibrosis	0.9	0.6	0.7	1.3	0.6‡
Pulmonary infarction	0	0	0	0	–
Cardiovascular
Angina/artery bypass	0.1	0	0	0.2	1‡
Myocardial infarction	0.1/0.1	0/−	0/−	0.2/0.2	–
Cardiomyopathy	0.9	0.3	0.7	1.3	0.21‡
Valvular disease	0.9	0.9	0	1.1	0.75‡
Pericarditis/pericardectomy	1.6	0.9	0.7	2.3	0.91‡
Peripheral vascular
Claudication for 6 months	0.3	0	0	0.5	0.67‡
Minor tissue loss	2.4/−	0.9/−	1.4/−	3.6/−	0.28‡
Significant tissue loss ever	0.7/0.3	0.3/−	0.7/0.7	0.9/0.4	0.11‡
Venous thrombus swelling, ulcer/stasis	2.5	1.7	0.7	3.4	0.81‡
Gastrointestinal
Bowel infarction or resection	1.3/0.3	1.1	1.4/0.7	1.4/0.4	0.88+
Mesenteric insufficiency	0.1	0.3	0	0	0.63‡
Peritonitis	0.2	0	0	0.4	0.63‡
Stricture/upper GI tract surgery ever	0.4	0.6	0	0.4	1‡
Pancreatic insufficiency	0.5	0.3	0	0.7	1‡
Musculoskeletal
Muscle atrophy/weakness	7.9	4	3.6	11.4	0.0056
Deforming/erosive arthritis	4.7	2.9	3.6	6.3	0.19
Osteoporosis with fracture	2.6	1.4	2.2	3.4	0.39‡
Avascular necrosis	2.7/1.0	0.6/−	10.8/5	2/0.4	0.055‡
Osteomyelitis	0.4	0	0.7	0.5	0.63‡
Ruptured tendons	0.1	0.3	0	0	0.41‡
Skin
Scarring chronic alopecia	9.1	11.7	6.5	8.2	0.06
Extensive scar/panniculum	1.7	1.4	1.4	2	0.82‡
Skin ulceration for 6 months	2.7	1.7	0	4.1	0.27
Premature gonadal failure	2	0.6	0.7	3.2	0.24‡
Diabetes	0.9	1.1	1.4	0.5	1‡
Malignancy	0.2	0.3	0.7	0	–

^*Values are the percentage of patients with Systemic Lupus International Collaborating Clinics/American College of Rheumatology Damage Index (SDI) item score of 1 or 2, or for end-stage renal disease a score of 3, unless indicated otherwise. Paired numbers are for items that could be counted twice.

UK = United Kingdom; CCHMC = Cincinnati Children’s Hospital Medical Center; PRINTO = Paediatric Rheumatology International Trials Organisation; GFR = glomerular filtration rate; GI = gastrointestinal.

^†From logistic models after adjusting for disease duration.

^‡From Fisher’s exact test.

Relationship between physician-rated damage severity and SDI summary scores.

Ratings of MD VAS damage were available for 1,792 of a total of 1,819 patient visits with SDI scores. Of these, 1,245 (69.4%) were without damage (SDI summary score = 0), and 547 (30.5%) had some damage (SDI summary score >0).

Damage-free patients (SDI summary score = 0) are expected to have an MD VAS damage of 0. However, physicians considered 24.2% of damage-free patients (301 of 1,245) as having some damage (MD VAS damage >0). As shown in Figure 2A, with SDI summary scores of 0, MD VAS damage ratings were <1 in 94% of the visits. Figure 2B provides an overview of the MD VAS damage and SDI summary scores throughout the range of observed values. Only 5.7% of patients (31 of 547) with SDI summary scores >0 were considered damage-free by their treating physician (MD VAS damage = 0).

Figure 2.

Systemic Lupus International Collaborating Clinics/American College of Rheumatology Damage Index (SDI) summary scores and physician rating of perceived damage severity in childhood-onset SLE. A, The relationship between physician-rated severity of damage (MD-VAS_damage) and the SDI summary score in patients with childhood-onset SLE is shown, with a focus on MD VAS damage ratings of 0, 1, or 2, showing ratings of up to 2 (range 0–10). Despite SDI scores of 1 or 2, some physicians provided damage severity ratings of 0. Conversely, only 76% of patients with SDI summary scores of 0 received the best possible MD VAS damage rating. B, The relationship between physician-rated severity of damage (MD-VAS_damage) and the SDI summary score in patients with childhood-onset SLE is shown, with SDI scores up to 12. With higher SDI summary scores, more commonly higher MD VAS damage ratings are assigned.

The correlation between SDI summary score and MD VAS damage overall was strong (Spearman’s r = 0.71, P < 0.0001), though when narrowed to include only visits with some damage (SDI score >0), the correlation was only moderate (Spearman’s r = 0.496, P < 0.0001). Correlations were similar when considering only the first or last visit for each patient, ranging from 0.66–0.72 for all SDI scores, versus 0.45–0.54 for only those visits with SDI summary scores >0.

In exploratory analysis using mixed-effects modeling, we aimed at identifying SDI items that importantly influence physician-rated damage severity. This analysis showed that only 4 of the 41 SDI items were significantly associated with the MD VAS damage. They were the SDI items pulmonary fibrosis, shrinking lung syndrome, chronic pericarditis, and extensive cutaneous scar. Of note, all of these SDI items had an overall frequency of <2% in the composite study cohort.

Consensus conference results.

Neither the SDI nor the proposed pediatric adaptation of the SDI (11) were considered adequate to measure damage severity in childhood-onset SLE. There was consensus (100%) that a separate measure or approach to capture damage severity of childhood-onset SLE was needed, and that the OMERACT filter 2.0 framework should be used for its development (83% agreement). Consensus was reached around a definition of childhood-onset SLE–associated damage and damage severity as a first step toward improving measurement of damage-related constructs.

Damage associated with childhood-onset SLE was defined as: “Impairment of anatomy or physiology that may be associated with scarring, may accumulate, and is not completely reversible. Damage may be caused by disease, adverse effects of medication, or associated comorbidity. In children this may lead to stunted cognitive, and physical development.” This definition had an 83% consensus.

Based on consensus (77% agreement), damage severity was defined as follows: “Severity of damage is measured by the organs involved, and the extent of anatomical and physiologic derangement as judged by the expected impact on mortality, degree of support required, activity limitation, restriction in social participation, and patient-centered quality of life.”

DISCUSSION

We examined the patterns and severity of disease damage in a large composite cohort of childhood-onset SLE patients, with the goal to appraise the ability of the SDI to capture disease damage and its severity. Based on detailed statistical analysis and expert consensus, the SDI was considered inadequate for quantifying the impact of disease damage in childhood-onset SLE.

Although childhood-onset SLE damage patterns have previously been described in detail (4,7,12), our analyses seem unique given the number of childhood-onset SLE patients included and the analysis of physician global assessments of damage severity. Damage patterns were largely similar across disease cohorts, with SDI items in the neuropsychiatric, renal, musculoskeletal, and skin systems most frequently encountered. Dissimilarities in damage patterns were identified for the frequency of chronic muscle atrophy or muscle weakness across cohorts. Reasons for these differences are unknown but could include differences in access to and use of medications, as well as differences in the relevant health care systems and demographics. Unfortunately, insufficient longitudinal data were available to identify the driving factors for the observed differences. Only 2 patients in our cohort encountered malignancies (type unknown), and this observation is in line with the malignancy risk based on an earlier epidemiologic study in childhood-onset SLE (13), supporting the representativeness of this composite cohort. Furthermore, the rarity of most SDI items in our cohort is in line with prior studies in adults (14).

Both SDI organ domain scores and total SDI summary scores increased with disease duration. Indeed, SDI summary scores were closely correlated with disease duration. This finding supports our prior research, where disease duration and cumulative disease activity were closely correlated with each other, but the cumulative burden of disease activity with childhood-onset SLE was the better predictor of SDI summary scores in a smaller childhood-onset SLE cohort (4). A close relationship between damage accrual, as measured by the SDI, and disease duration supports the construct validity of the SDI for measuring damage.

The SDI summary score is the simple sum of its mostly unweighted SDI item scores. Thus, one might expect that a higher score corresponds to more severe damage, especially because there are 7 high-impact items that receive an additional score (myocardial infarction, stroke, end-stage renal disease, avascular necrosis, malignancy, significant tissue loss, and gastrointestinal infarction/resection) if occurring repeatedly. While true item weightings have been incorporated in other scales used for SLE and childhood-onset SLE (4,15), our findings imply that more sophisticated item weightings will not improve the ability of the SDI to capture childhood-onset SLE–associated damage severity. This inability is because there were only statistically significant associations between the MD VAS damage summary score and a few, rarely endorsed, SDI items. This observation is similar to that of earlier studies using other statistical approaches, which also found little value in item weightings to improve the SDI’s ability to reflect damage severity, or mortality (2,4,6).

Although disease damage accumulated with childhood-onset SLE was substantial, many of the SDI items were rarely scored. Indeed, nearly half of SDI items were encountered in <1% of childhood-onset SLE patients. Although eliminating items with extremely low prevalence might be tempting, to improve the feasibility of the scale when used in childhood-onset SLE, we do not think this practice is advisable, as item reduction will not improve the construct validity of the SDI, nor its ability to capture damage severity. Further, SLE and childhood-onset SLE are highly heterogeneous in their phenotypes and certain SDI items, such as myocardial infarctions, are known to occur more commonly with higher age and longer disease durations than those captured by our study. There is also broad agreement that the inclusion of rarely scored items is warranted when estimating damage with other multisystem diseases, such as inflammatory immune-mediated myositis and vasculitis (16,17).

If the SDI were a good measure of childhood-onset SLE–associated damage severity, children with SDI summary scores of 0 (best possible value) should consistently have an MD VAS damage rating of 0 (best possible value). However, over 20% of children without childhood-onset SLE–associated damage received an MD VAS damage rating >0. While the exact reasons for this observation are unknown, it could simply be a reflection of so-called end-scale aversion; raters are known to avoid providing the best or worse possible rating on a VAS (18,19). Actually, almost all MD VAS damage ratings with SDI summary scores of 0 were in the range of 0 to <1.

Another explanation for MD VAS damage ratings exceeding 0 in patients without damage may be that certain pediatric-specific damage is not considered in the SDI currently. Indeed, there is consensus among pediatric rheumatologists that pubertal development and growth are important aspects that deserve consideration when measuring childhood-onset SLE–associated damage (8). This belief is reflected in the proposal of a pediatric version of the SDI, where 2 items are added to the traditional SDI (11), namely reduced growth and delayed development. A problem with these childhood-onset SLE–specific items might be that both items can resolve, which could be perceived to violate the concept of irreversibility of disease damage as defined by the SDI. Pediatric experts involved in the delineation of the childhood-onset SLE damage core set considered this issue acceptable, given the more pronounced tissue regeneration in children (8). An example to be considered is the resolution of some bone erosions in pediatrics.

The concept of reversibility of some damage is now clearly stated in the consensus definition of childhood-onset SLE–associated damage. Notably, some of the current SDI items can also resolve clinically, such as nephrotic-range proteinuria or seizures, but the resultant SDI item scores are maintained (2). Whether it would be more sensible to reduce the summary score when growth and development delays have resolved, or to continue considering them in a damage summary score, will need further evaluation, should the validation of the pediatric SDI be pursued.

For the SDI to be considered a robust measure of childhood-onset SLE–associated damage severity, one would expect a strong correlation between the SDI summary score and the MD VAS damage rating. However, our data suggest that this correlation is not the case, which is in line with our prior research (6). Indeed, the lack of a strong association of damage severity with commonly encountered SDI items likely reflects the SDI originators’ caution that the index provides only an enumeration of damage items, rather than quantifying severity (1). Hence, one might suggest the use of the term “summary count,” rather than “summary score,” to emphasize the distinction in future research. Thus, presenting means and SDs of the SDI summary score or using the SDI summary score as a continuous variable in statistical modeling seems inadvisable.

The limitation of the SDI to measure damage severity is not surprising, because SDI items vary considerably in their impact on patient function and on the need for medical interventions. For example, a stroke resulting in hemiparesis will likely have greater impact on patient function than a cataract, despite the equal summary score contribution of each item. Indeed, physician perception of damage severity is likely influenced by both the type of damage and the resultant prognosis and/or health care utilization. This fact might indicate that there is a need for Likert scaling of SDI items to better capture the extent or impact of damage encountered. For example, a stroke that does not result in long-term clinical deficits would receive a lower item score than a stroke causing mild localized paresis, which in turn would be scored less than a stroke that renders the patient unable to walk or to speak.

Perspective is also a key factor when estimating the severity of childhood-onset SLE damage. This necessity can be exemplified by scarring alopecia, which has an extremely limited impact on mortality, but could be devastating from a patient’s viewpoint. These considerations are further complicated by the known resilience of children (11).

The shortcomings of the SDI were highlighted by deliberations of childhood-onset SLE experts during the recent consensus conference. In accordance with the OMERACT filter 2.0 (10), the importance of perspective when measuring the severity or impact of childhood-onset SLE–associated damage is reflected in the newly developed provisional consensus definition of damage severity: “As judged by the expected impact on mortality, degree of support required, activity limitation, restriction in social participation, and patient-centered quality of life.”

Limitations of our study include the lack of physician damage severity ratings from all patients included in the composite cohort. Use of physician global assessments may be considered problematic in itself (20), with potential interrater and intrarater variation, which we were unable to assess. Nonetheless, over 1,200 MD VAS damage ratings were available for analysis, and VAS ratings are widely used in medical research in general and rheumatology in particular.

Taken together, the analysis of a large childhood-onset SLE data set confirms the construct validity of the SDI as a valid measure to count damage events in children and adolescents with childhood-onset SLE. The SDI in its current form is ill suited to accurately quantify the severity of childhood-onset SLE–associated damage, even if item weightings were revisited in an effort to improve the algorithm used to calculate the SDI summary score. We propose an international effort, supported by the relevant professional organizations, to capture the severity of impact of disease damage of patients with SLE from childhood through adulthood.

Significance & Innovations.

• In childhood-onset systemic lupus erythematosus (SLE), damage is common, and occurs most frequently in the renal, cutaneous, neuropsychiatric, and musculoskeletal organ systems.

• By design, the Systemic Lupus International Collaborating Clinics/American College of Rheumatology Damage Index (SDI) seeks to provide an enumeration of the types of damage present in childhood-onset SLE, rather than directly measuring damage severity. Despite this fact, the SDI summary score is often used as a stand-alone continuous outcome measure in research.

• An international group of pediatric rheumatologists and nephrologists experienced in childhood-onset SLE reached unanimous consensus that a new measure or approach is needed to better capture damage severity in childhood-onset SLE.