Risk factors for damage in childhood-onset systemic lupus erythematosus: Cumulative disease activity and medication use predict disease damage




The Systemic Lupus International Collaborating Clinics/American College of Rheumatology (SLICC/ACR) Damage Index measures damage in adult patients with systemic lupus erythematosus (SLE), but its usefulness in patients with childhood-onset SLE has not been examined. This study was conducted to evaluate the sensibility of the SLICC/ACR Damage Index, to investigate how cumulative disease activity is related to damage in childhood-onset SLE, and to identify other risk factors for damage in childhood-onset SLE.


Disease activity and damage in 66 patients with newly diagnosed childhood-onset SLE were assessed retrospectively, and information on potential risk factors for damage (age, race, sex, medications, duration of disease, hypertension, body mass index, antiphospholipid antibodies, kidney disease, acute thrombocytopenia) was obtained. In addition, a group of physicians was surveyed to establish the sensibility of the SLICC/ACR Damage Index in childhood-onset SLE.


The SLICC/ACR Damage Index was found to have face, content, and construct validity when used in children. The mean SLICC/ACR Damage Index score of the patients was 1.76 (mean followup 3.3 years). Cumulative disease activity over time was the single best predictor of damage (R2 = 0.30). Other, possibly important risk factors for damage were corticosteroid treatment, the presence of antiphospholipid antibodies, and acute thrombocytopenia. It was determined that immunosuppressive agents may be protective.


The SLICC/ACR Damage Index, though useful in childhood-onset SLE, may benefit from the introduction of weightings and redefinition of some of the items. Ongoing disease activity leads to disease damage, and treatment should be prompt. Prolonged use of high-dose corticosteroids may further increase damage, but use of immunosuppressive agents may protect against disease damage; this latter finding may have potential implications for the treatment of childhood-onset SLE and deserves further study. The relationship between disease activity and concomitant use of medication also requires further investigation.

Systemic lupus erythematosus (SLE) is a multisystem, inflammatory, autoimmune disease, the course of which is characterized by periods of flare and remission (1). The degree of inflammation observed in patients with SLE varies throughout the course of the disease. Inflammation, as well as the side effects and complications of treatment, can result in irreversible tissue damage and may even lead to premature death. Terms such as disease activity and disease damage (1) are commonly used to describe, respectively, the varying pattern of inflammatory intensity and the permanent damage observed in SLE patients.

Neither disease activity nor disease damage in lupus can be measured directly with any individual laboratory value or clinical sign. Thus, indices of disease activity and damage have been developed for use in adults with SLE (2–5). Because the clinical and laboratory features associated with childhood-onset SLE, i.e., disease onset prior to age 18 years, are different from those associated with adult SLE (6–14), all disease indices that have been developed and tested for use in adults require reevaluation prior to use in children (14). The Systemic Lupus Erythematosus Disease Activity Index (SLEDAI [ 5]) is commonly used in North America to measure disease activity in adults and has been validated for use in patients with childhood-onset SLE (14). The Systemic Lupus International Collaborating Clinics/American College of Rheumatology (SLICC/ACR) Damage Index was developed to measure nonreversible damage (15, 16) in adults with SLE, but its usefulness in patients with childhood-onset SLE has not been investigated.

Disease damage, morbidity, and mortality in childhood-onset SLE have previously been linked to predictors (risk factors) that are not considered in the SLEDAI. Such risk factors include young age at diagnosis (9, 17, 18), male sex (5, 19–22), and nonwhite (black, Asian, and Hispanic) ethnicity (7). However, an association between any of these risk factors and a poor prognosis is controversial (5, 13). Acute thrombocytopenia has also been associated with a poor prognosis; it has been hypothesized that a sudden drop in platelet count is a marker for acute, very aggressive disease as opposed to chronic thrombocytopenia (23–26). Patients with antiphospholipid antibodies (aPL) (anticardiolipin antibodies [aCL] or lupus anticoagulant [LAC]) are at increased risk for recurrent thrombosis, stroke, thrombocytopenia, and cardiac disease. Therefore, the presence of aPL may predispose to disease damage (18, 27, 28). Hypertension is a feature of childhood-onset SLE that is observed especially in patients treated with high-dose corticosteroids and in those with severe lupus nephritis (11, 21). The clinical feature most consistently associated with a poor prognosis is diffuse proliferative glomerulonephritis (type IV nephritis according to World Health Organization [WHO] classification), especially when this type of nephritis is associated with hypertension (5, 11, 21, 25, 26, 29, 30).

Deciding whether a child who has been diagnosed with childhood-onset SLE has a greater or lesser degree of disease activity and determining how to avoid disease damage are essential for patient management. Although an association between disease activity and permanent damage makes sense intuitively, this relationship has not been well investigated. It is unclear to what degree any risk factor, alone or in combination with ongoing disease activity, affects the damage observed in childhood-onset SLE.

Our primary objective in the present study was to determine whether cumulative disease activity over time leads to disease damage in children with SLE. Secondary objectives were to examine the sensibility (face and content validity) of the SLICC/ACR Damage Index when used in children and to identify any predictors of disease damage other than disease activity.


Instruments.SLEDAI. The SLEDAI is a disease-specific scale of disease activity developed for use in adults with SLE. It measures potentially reversible manifestations of the underlying inflammatory disease process (5). The scale consists of 24 “weighted” attributes grouped into 9 domains (organ systems), as follows: weighting of 8 for central nervous system and vascular, 4 for renal and musculoskeletal, 2 for serosal, dermal, and immunologic, and 1 for constitutional and hematologic. If, during a 10-day period prior to the assessment, a patient manifests a clinical variable, then the corresponding weighted score is assigned. The final score comprises the sum of all weighted attribute scores. The SLEDAI has a theoretically possible range of 0 to 105, with 0 being no disease activity. This instrument has been validated for use in patients with childhood-onset SLE and is sensitive to clinically important change in disease activity when used in a pediatric population (14).

SLICC/ACR Damage Index. Earlier definitions of disease damage in SLE were confined to damage caused by inflammation of the underlying disease itself (16). During development of the SLICC/ACR Damage Index, the only currently available instrument for measuring disease damage in SLE, it was recognized that damage from inflammation cannot be clearly distinguished from damage secondary to medication side effects or other comorbid conditions. Therefore, both disease-specific and non–disease-specific damage are included in the SLICC/ACR Damage Index. This instrument, which was developed through a group process, quantifies nonreversible cumulative damage that has occurred since SLE was diagnosed.

According to the SLICC/ACR Damage Index, damage is considered nonreversible if any given item was present for at least 6 months continuously (16). Therefore, even if an item was present for longer than 6 months then subsequently resolved (i.e., was actually reversible), the patient is still assigned the score. This scoring method is used because the developers of the tool believe that “persistent inflammation for at least 6 months would cause some tissue injury, resulting in damage” (16).

The SLICC/ACR Damage Index is unweighted, and a definition is provided for all 41 items (31). A score of 0 is assigned to patients with no nonreversible damage, and the maximum possible score is 47. Damage evaluated by this index is grouped into 12 different domains and assigned maximum scores as follows: 1 for premature gonadal failure and diabetes mellitus, 2 for ocular and malignancy, 3 for renal and skin, 4 for peripheral vascular, 5 for pulmonary, 6 for neuropsychiatric and cardiovascular, and 7 for gastrointestinal and musculoskeletal.

Patients. The hospital charts of all patients who were newly diagnosed with childhood-onset SLE between January 1990 and June 1998 and followed up at the Lupus Clinic at The Hospital for Sick Children (HSC) were reviewed. All patients fulfilled the ACR classification criteria for SLE (31) and either received a diagnosis of childhood-onset SLE at the HSC or began regular followup at the Lupus Clinic at the HSC within 1 year postdiagnosis. The minimum followup time of patients included in the study was 6 months. Disease activity was measured with the SLEDAI (10) at the time of all clinic visits and hospital admissions. All data necessary to complete the SLEDAI were extracted from the standardized clinic forms used at the HSC and from laboratory databases that stored the results of standard laboratory tests. Menses of female patients were scored only as being either present or absent. The Tanner stage and information regarding the concomitant use of systemic contraceptive drugs were not regularly recorded in the charts.

The SLICC/ACR Damage Index scores were determined at the end of the study period. Additional data were obtained from the charts to assess the impact of the following potential risk factors: 1) cumulative corticosteroid dose, 2) length of treatment with high-dose corticosteroids (≥0.3 mg/kg/day or ≥20 mg/day), 3) number of pulse treatments with methylprednisolone, 4) length of treatment with cyclophosphamide, 5) length of treatment with azathioprine, cyclosporin A, or methotrexate, 6) duration of hypertension, 7) mean body mass index (kg/m2) during the study period, 8) presence of LAC, 9) presence of elevated levels of anticardiolipin antibodies (aCL), 10) prior episodes of acute thrombocytopenia (platelet count <100,000 × 109/mm3 for <3 months), 11) type of kidney disease according to WHO classification of lupus nephritis, 12) sex, 13) race (white/nonwhite), 14) age at onset of disease, and 15) absolute duration of disease.

Face and content validity of the SLICC/ACR Damage Index. To investigate the face and content validity of the SLICC/ACR Damage Index in children, the instrument itself was given to 7 local pediatric rheumatologists, 2 pediatric nephrologists with interest in childhood-onset SLE, and 5 pediatric rheumatology fellows, all of whom were asked to assess and review it. In the past, the developers of the SLICC/ACR Damage Index used a similar approach to establish the sensibility of the SLICC/ACR Damage Index in adults (16).

Measurement of cumulative disease activity over time. Cumulative disease activity over time was determined by calculating the area under the curve (AUC) of the serial measurements of disease activity using the trapezoidal rule. The algorithm used to calculate the AUC of ongoing disease activity as measured by the SLEDAI (AUC-S) was as follows (33):

equation image

where AUC-S(m) = the AUC of the cumulative SLEDAI scores up to time m, n = the number of assessments (n = 1, 2, … m), and SLEDAI(n) = the SLEDAI score at time n.

Statistical analysis to investigate the relationship between cumulative disease activity and disease damage. Linear modeling was used to describe the relationship between cumulative disease activity over time (AUC-S) and disease damage. The linear model was examined for its fit, and regression diagnostics were performed to evaluate statistical performance (34–37).

Statistical analysis to investigate the relationship between cumulative disease activity and disease damage under consideration of other risk factors. Fifteen proposed risk factors were examined by univariate analysis for their statistical importance. Different initial groups of covariates were formed to avoid statistical problems (extreme colinearity) in subsequently performed analyses. Clinical experience, regression diagnostics, and a uniform selection algorithm (34, 35, 38–40) were used to further assess the quality (statistical performance) and usefulness of these multivariable models in medical practice. Microsoft Excel 97 spreadsheets (Redmond, WA) and the statistical package SAS, version 6.12 (Cary, NC) were used for the statistical analysis.


Face validity of the SLICC/ACR Damage Index for childhood-onset SLE. Fourteen questionnaires for assessing the sensibility of the SLICC/ACR Damage Index (16) were distributed, 2 of which were not returned (response rate 86%). Therefore, 12 completed questionnaires were available for evaluation. Eleven (92%) of the 12 physicians indicated that the Index was a sensible tool for measuring disease damage in patients with childhood-onset SLE. One physician noted that the instrument in its current form is not useful in children, because it is unweighted. For the same reason, 3 other physicians anticipated problems using the Index in children. It was also expected that measuring cognitive deficits scored by the SLICC/ACR Damage Index would be difficult. Comments were made regarding the definition of nonreversibility used by the Index (i.e., the presence of an item for at least 6 months continuously), and it was thought that a 6-month period of time might be too short. Two physicians raised concerns about the definition and measurement of premature gonadal failure. Five physicians suggested redefining the item “cataracts,” so that only vision-compromising cataracts would contribute to the score of a patient. Additional comments were directed toward the definition of proteinuria. It was suggested that nephrotic-range proteinuria should be redefined as “proteinuria ≥50 mg/kg/day.”

Description of patients and therapies. Between January 1990 and June 1998, 66 patients with newly diagnosed childhood-onset SLE were seen at the HSC. They were followed up for an average of 3.3 years (for a total of 219 patient-years) (Table 1). Thirty-four patients were hypertensive on at least 2 consecutive clinic visits, and 19 children had hypertension that persisted for >1 year. Kidney biopsy specimens were available from 43 patients (65% of all patients) with clinically suspected renal disease (WHO class II glomerulonephritis in 23%, class III in 25%, class IV in 40%, and class V in 12%). Ten of the 38 aPL-positive patients developed symptoms compatible with aPL syndrome. Acute thrombocytopenia was observed in 23 patients. During the study period, high-dose corticosteroids were intermittently required by 63 of the 66 patients, pulse treatments with methylprednisolone were given to 21 patients, and 43 patients received immunosuppressive agents.

Table 1. Demographic data on the cohort
  • *

    Only patients whose parents were both white were considered as being white for the purpose of this analysis. Nonwhite patients include 11 who were black, 11 Chinese, 1 Japanese, 5 Indian, 8 South Asian, 4 Asian Pacific Islanders, and 8 other Asian.

Age in years, mean ± SD/median (range)
 At time of diagnosis12.9 ± 2.49/13.1 (4–17)
 At end of followup16.0 ± 2.63/16.6 (9.3–19.9)
Followup in years, mean ± SD/median (range)3.3 ± 2.04/3.2 (0.5–7.9)
No. female/no. male56/10
No. white/no. nonwhite*18/48

SLICC/ACR Damage Index scores. At the end of the followup period, 26 of the 66 patients had a SLICC/ACR Damage Index score of 0. The other 40 patients were assigned Index scores as follows: 16 patients had a score of 1, 15 patients had a score of 2 or 3, 5 patients had a score of 4 or 5, 1 patient had a score of 8, and 3 patients had a score of 11 or 12. The combined total SLICC/ACR Damage Index score of the study group was 116, and the mean ± SD score was 1.76 ± 2.64 (range 0–12, median 2).

Damage occurred primarily in the musculoskeletal, ocular, neuropsychiatric, and renal domains (Table 2), with musculoskeletal damage being most common (34 [29%] of 116 scores). Avascular necrosis, usually occurring early in the course of the disease, was diagnosed in 15 patients (27 [13%] of 116 scores). Ocular damage accounted for 33 scores, and small cataracts accounted for 29 of these 33 scores. Seven patients (11%) were scored as having neuropsychiatric damage which accounted for 17% of the total damage (20 of 116 scores). Damage scores due to cognitive impairment of at least 6 months' duration were based mostly on the presence of major school problems or reports from family members. Renal damage occurred in 6 patients (11 of 116 scores), representing 9.5% of the total damage. Both patients in whom renal failure ultimately developed had WHO class IV lupus nephritis with concomitant features of class V nephritis, on renal biopsy. One patient who was undergoing dialysis died because of multiorgan failure and sepsis.

Table 2. Systemic Lupus International Collaborating Clinics/American College of Rheumatology (SLICC/ACR) Damage Index scores at time of last assessment*
Domain, itemScore
  • *

    By the end of the followup period, 40 of 66 patients had SLICC/ACR Damage Index scores >0; the cohort accumulated a total SLICC/ACR Damage Index score of 116.

1. Ocular (n = 29)
 Any cataract29
 Retinal changes or optic atrophy4
2. Neuropsychiatric (n = 7)
 Cognitive impairment or major psychosis6
 Seizures requiring therapy for 6 months4
 Cerebrovascular accident (CVA) (score 2 if >1)4 (all multipleCVAs, score 8)
 Cranial or peripheral neuropathy2
 Transverse myelitis
3. Renal (n = 6)
 Estimated or measured glomerular filtration rate <50%2
 Proteinuria ≥3.5 gm/24 hours3
  or end-stage renal failure (regardless of dialysis or transplantation)2 (score 6)
4. Pulmonary (n = 2)
 Pulmonary hypertension
 Pulmonary fibrosis2
 Shrinking lung
 Pleural fibrosis
 Pulmonary infarction or resection not for malignancy
5. Cardiovascular (n = 1)
 Angina or coronary artery bypass 
 Myocardial infarct (score 2 if >1)
 Cardiomyopathy (ventricular dysfunction)
 Valvular change (murmur)1
 Pericarditis for 6 months or pericardiectomy
6.Peripheral vascular (n = 2) 
 Claudication for 6 months
 Venous emboli with swelling, ulceration or venous stasis
 Minor tissue loss (pulp space)1
 Significant tissue loss (score 2 if >1)1 (score 2)
7. Gastrointestinal (n = 3)
 Infarction or resection of bowel, spleen, liver, or gall bladder (score 2 if >1 site)1 (score 2)
 Mesenteric insufficiency2
 Chronic peritonitis1
 Stricture or upper gastrointestinal surgery ever
 Pancreatic insufficiency (enzyme replacement or with pseudocyst)
8. Musculoskeletal (n = 17)
 Atrophy or weakness1
 Deforming or erosive arthritis2
 Osteoporosis with fracture or vertebral collapse3
 Avascular necrosis (score 2 if >1)15 (score 27)
 Ruptured tendon
9. Skin (n = 4)
 Extensive scarring or panniculum other than pulp space and scalp2
 Skin ulceration (excluding thrombosis for >6 months)3
10. Premature gonadal failure0
11. Diabetes (regardless of treatment) (n = 2)2
12. Malignancy (score 2 if >1 site)0

Cumulative disease activity as a predictor of damage. Although the AUC-S was generally a very good predictor of damage (R2 = 0.30, slope = 22.3, SD of slope = 4.22, P value of slope <0.0001, intercept = 0.125, P value of intercept not significant [NS]), the damage in 9 of the 66 patients was not well predicted by the linear model (outlying observations). No special clinical, laboratory, or demographic features of these 9 patients distinguished them from the remaining study population. An exploratory analysis was performed to examine whether cumulative renal disease activity alone (AUC of SLEDAI scores from the renal domain) and ongoing very high disease activity over time (AUC of SLEDAI scores >24) were better predictors of damage than the AUC-S. However, these special measures of cumulative disease activity were not superior to total cumulative disease activity over time (AUC-S) for estimating disease damage.

Other risk factors for disease damage. Considering the limited sample size in this study, all the multivariable models must be regarded as generating hypotheses that will need further testing.

Variable reduction.Besides cumulative disease activity over time (AUC-S), we considered 15 other potential risk factors. However, number of pulse treatments with methylprednisolone, average body mass index during the followup period, and race and sex were subsequently excluded from testing in multivariate analysis, because these factors were not correlated with either damage or cumulative disease activity (Table 3).

Table 3. Predictors of disease damage in childhood-onset systemic lupus erythematosus
  • *

    Tested in linear and multivariable analysis.

  • As identified in multivariable analysis.

  • Examined but found to be unrelated to disease damage or cumulative disease activity.

  • §

    Statistically important by multivariable analysis.

Important risk factor*
 Cumulative disease activity over time
Likely important risk factors
 Cumulative steroid dose
 Time taking high-dose steroids (daily dosage ≥0.3 mg/kg or  ≥20 mg)
 Presence of antiphospholipid antibodies
 Acute thrombocytopenia
Likely not important predictors
Pulse treatments with methylprednisolone
 Age at diagnosis
 Disease duration
 Body mass index
 Kidney disease
 Duration of hypertension
Possibly important protective factor§
 Duration of therapy with immunosuppressive agents

Multivariable models of disease damage. Cumulative disease activity over time (AUC-S) and relevant risk factors (whose importance was established in the univariate analysis) were used for the regression analysis and multivariable modeling. As expected, all multivariable models had a higher explanatory power to predict damage than did the linear model, with AUC-S as the only predictor (R2 = 50–57%, all intercepts <14.0, P value of all intercepts NS, all standardized β coefficients >0.2, P value of all standardized β coefficients <0.0001). Important risk factors for development of damage were length of treatment with high-dose corticosteroids, total cumulative dose of corticosteroids, presence of aPL, and history of acute thrombocytopenia. It was demonstrated that immunosuppressive agents likely have a protective effect against damage (standardized β coefficients <−0.2, P value of all standardized β coefficients <0.0001). Conversely, duration of hypertension, type of kidney disease, age at diagnosis, and duration of followup were not important predictors of damage in childhood-onset SLE (Table 3).


The main objective of this study was to investigate the relationship between disease activity (as measured by the SLEDAI) and disease damage (as scored by the SLICC/ACR Damage Index) in childhood-onset SLE. Therefore, cumulative disease activity over time in 66 patients with newly diagnosed childhood-onset SLE was measured and found to be, by itself, a highly significant predictor of the development of disease damage. We also found that corticosteroid therapy, the presence of aPL, and acute thrombocytopenia predispose to damage, whereas treatment with immunosuppressive agents might protect against damage. The relationship between cumulative disease activity over time and damage in childhood-onset SLE further confirms the SLICC/ACR Damage Index for use in children and supports its construct validity in this patient population.

Results of the evaluation of the face and content validity of the SLICC/ACR Damage Index suggest that it may benefit from the introduction of weighting of its items. Some of the physicians who participated stated that it did not seem sensible to give equal scores for a cataract and a stroke or a malignancy. Reevaluation of the need for weighting of the SLICC/ACR Damage Index items may be warranted, especially because only one study has examined this issue in the past (41). We examined the effect of weightings—similar to those used in the SLEDAI—in a secondary analysis (data not shown), but this approach did not improve the predictive power of cumulative disease activity over time.

In addition, the SLICC/ACR Damage Index may benefit from changes in some item definitions for use in childhood-onset SLE. A clearer and more specific definition of “cognitive impairment” should be considered. For instance, a certain score on a validated psychometric instrument might be chosen to identify cognitive impairment that qualifies for SLICC/ACR Damage Index scores. The definition of “premature gonadal failure” based on secondary amenorrhea alone may require modification for childhood-onset SLE. In adolescents, irregular menstrual cycles are part of the normal body maturation process and, as such, “secondary amenorrhea” may not be a good surrogate for premature gonadal failure in childhood-onset SLE. Hormonal studies are likely more appropriate for establishing such a diagnosis in this age group. In addition, concomitant use of systemic contraceptives may have to be considered.

The definition of nephrotic-range proteinuria currently used in the SLICC/ACR Damage Index (proteinuria ≥3.5 gm/day) should be changed to the weight-based definition commonly used in pediatrics. Of note is the fact that neuropsychiatric scores in 2 patients were assigned based on treatment with anticonvulsive medication that commenced after hypertensive seizures. Therefore, these scores were not given for neuropsychiatric damage in childhood-onset SLE, per se, but for complications of severe renal disease. As such, the definition of this item may require modification to more accurately reflect the neuropsychiatric disease damage.

Evaluation of the face validity of the SLICC/ACR Damage Index for use in childhood-onset SLE was based on a relatively small number of ratings, some of which were performed by pediatric rheumatology trainees. Thus, all suggested changes of item definitions must be rediscussed by a larger group of lupologists and tested in future studies.

The children in this study accumulated disease damage more rapidly than do adults. The mean SLICC/ACR Damage Index score was 1.76 after an average disease duration of only 3.3 years, which is more than twice the mean damage score reported in adult cohorts 5 years postdiagnosis (42). This finding is consistent with prior reports suggesting that childhood-onset SLE has a more aggressive course than that of adult SLE (19, 21, 25), and that treatment with corticosteroids and immunosuppressive agents is required more often in children.

Use of corticosteroids appeared to be an important risk factor for damage in our patients. Although pulse treatment with methylprednisolone has previously been associated with the development of avascular necrosis (18, 28, 43, 44), the number of pulse treatments with methylprednisolone was not related to the damage scores in this study. Although this lack of a relationship may have been attributable to the small sample size, pulse treatments with methylprednisolone were used with equal frequency in patients who did and those who did not develop avascular necrosis. In our cohort, it appeared that pulse therapy with methylprednisolone was not the primary risk factor for avascular necrosis. Our results suggest that long-term use of high-dose corticosteroids, rather than pulse therapy with methylprednisolone, contributes to disease damage. If this is truly the case, then childhood-onset SLE patients might benefit from pulse treatment with methylprednisolone, because it allows more rapid tapering of the daily steroid dosage without the same risk of damage.

Both cumulative corticosteroid dose and long-term treatment with high-dose corticosteroids were identified as potentially important risk factors. These 2 measures of corticosteroid use were highly correlated. In the past, the introduction of corticosteroids for treatment of childhood-onset SLE led to a significantly improved prognosis for this disease. Corticosteroids are known to contribute to the development of damage in adults with SLE (45), and they also seem to increase disease damage in patients with childhood-onset SLE. This may indicate that the use of corticosteroids in childhood-onset SLE should be refined. More judicious dosing and rapid tapering of daily oral corticosteroids, facilitated by using intravenous pulse methylprednisolone therapy and/or immunosuppressive agents, may possibly further increase the net benefits of corticosteroid therapy for children with SLE.

Currently, cyclophosphamide and other immunosuppressive agents are used mainly for the treatment of severe kidney and neuropsychiatric disease in patients with childhood-onset SLE. In adults with SLE, common side effects of immunosuppressive therapy are infection, leukopenia, and gonadal failure. These adverse effects were not observed in any of the childhood-onset SLE patients in this study who were treated with immunosuppressive agents. Treatment with immunosuppressive agents seemed to protect against development of irreversible disease damage. This benefit may be attributable to the steroid-sparing effects of immunosuppressive agents or could also be due to more effective control of disease activity and flares than would be possible with corticosteroids alone.

Nonwhite race has frequently been associated (13, 25, 46, 47) with a poor prognosis for patients with SLE. In this study, race was not a risk factor for development of disease damage. Also, nonwhite children did not have significantly higher cumulative disease activity compared with white children, nor did they develop more damage. It was previously suggested that the increased mortality of black SLE patients may be based on poorer access to health care among this group (48, 49). Given the structure of the Canadian health care system and its universal health care coverage, access to health care is not an issue at the HSC. Thus, results of this study support previous findings that decreased access to health care rather than race contributed to the poor prognosis for nonwhite SLE populations. However, the relatively short study period, the small study population, as well as the ethnic diversity of the cohort may have precluded identification of racial risk factors (25) in this study.

A potential limitation to our study is the fact that it was carried out at a single center. It is known that treatment of SLE (and likely childhood-onset SLE) varies between centers (50). Thus, the relative importance of the predictors identified in this study may change in other childhood-onset SLE cohorts. In general, however, corticosteroids as well as immunosuppressive agents are the cornerstones of treatment of childhood-onset SLE worldwide (51). Another limitation to this study is the possibility that the assessment of secondary amenorrhea was incomplete, because the standardized clinic forms used for the patient assessments inquired only about the presence or absence of menses without clear documentation of the relevant time frames. Thus, prolonged secondary amenorrhea that should have been scored in the SLICC/ACR Damage Index may have been missed. It is unlikely that this occurred often enough to have made a difference in our results; at the HSC, none of the childhood-onset SLE patients who completed treatment with cyclophosphamide but were not included in the study because of the chosen time frame reported persistent amenorrhea (Silverman ED: personal communication).

Estimates of disease activity and the probability of damage guide treatment decisions in childhood-onset SLE. The ultimate goal of therapy is to improve the prognosis of patients with childhood-onset SLE by preventing disease damage. This is the first study that systematically investigated the relationship between disease activity and disease damage in childhood-onset SLE. Our results support the concept that cumulative disease activity over time results in damage, and, therefore, rapid control of disease activity is desirable to minimize disease damage. Early introduction of therapy with immunosuppressive agents might improve the outcome in patients with childhood-onset SLE by preventing permanent damage. This study did not allow an estimation of the exact impact of corticosteroids on increased disease damage. However, our results strongly suggest that corticosteroids, though highly effective for controlling disease features, can be an important cause of damage. Patients may benefit from more judicious use of corticosteroids with refined tapering schedules and early introduction of steroid-sparing therapies.