SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. REFERENCES

Objective

Studies of adults with systemic lupus erythematosus (SLE) have frequently demonstrated the presence of decreased bone mineral density (BMD). However, there have been few investigations in pediatric patients to date. This study was undertaken to determine the prevalence of low BMD in patients with juvenile SLE and to identify associated risk factors.

Methods

We studied 64 consecutive patients with juvenile SLE in whom routine dual x-ray absorptiometry (DXA) scanning was performed. Lumbar spine osteopenia was defined as a BMD Z score of <−1 and ≥−2.5, and osteoporosis as a BMD Z score of <−2.5. Decreased hip BMD was defined as a value of <80%. Data on disease activity, quality of life, disease-related damage, sex, ethnicity, body mass index, age at diagnosis, age at DXA, medication use and duration, clinical features, and puberty status were collected at the time of DXA.

Results

Lumbar spine osteopenia was seen in 24 patients (37.5%) and osteoporosis in 13 (20.3%). Decreased hip BMD was present in 12 patients (18.8%). By univariate analysis, osteopenia was significantly correlated with age, disease duration, duration of corticosteroid use, cumulative corticosteroid dose, azathioprine use, cyclophosphamide use, lupus nephritis, and damage. Two additional variables, mycophenolate mofetil use and class III–IV nephritis, were associated with osteoporosis. Abnormal hip BMD was associated with disease duration, duration of corticosteroid use, and cumulative corticosteroid dose. By multivariate analysis, only disease duration remained in the model for osteoporosis and abnormal hip BMD, while cumulative corticosteroid dose was the variable associated with osteopenia.

Conclusion

These results indicate that osteopenia and osteoporosis are common in juvenile SLE and are associated more closely with increased disease duration than with cumulative corticosteroid dose.

Juvenile systemic lupus erythematosus is a multisystem autoimmune condition characterized by chronic inflammation and by the presence of autoantibodies. Although patients with juvenile SLE comprise only 15–20% of all SLE patients, it has been suggested that they have more severe disease than adults with SLE, necessitating more frequent use of high-dose corticosteroids (1). Despite the more severe disease course of juvenile SLE, however, there has been significant improvement in long-term mortality. Therefore, more studies have focused on long-term morbidities, specifically, premature atherosclerosis and osteoporosis leading to insufficiency fractures (2–4).

Osteoporosis, an illness generally associated with postmenopausal women, is characterized by loss of both bone mass and microarchitectural integrity (5). One crucial determinate in the development of osteoporosis is the acquisition of appropriate peak mass in late adolescence and early adulthood (6). A failure to achieve adolescent peak bone mass may be associated with premature osteoporosis and increased risk of fracture. The development of peak bone mass is the result of interactions between nutritional factors including calcium intake, environmental exposures, physical activity, and medications (7, 8). Of particular concern with regard to children with chronic rheumatic diseases are the detrimental effects of chronic inflammation and corticosteroid use (9–13). These concerns are particularly important in juvenile SLE, since these patients tend to have severe chronic inflammation and frequently receive prolonged courses of high-dose corticosteroid therapy. Despite these concerns, there have been few studies examining osteoporosis and the development of fragility fractures in patients with juvenile SLE (2, 14–16). The aim of the present study was to identify risk factors associated with osteoporosis in this patient population.

PATIENTS AND METHODS

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. REFERENCES

Patients.

In January 2001, dual x-ray absorptiometry (DXA) assessment became routine in the pediatric SLE clinic at the Hospital for Sick Children. Sixty-four consecutive patients who fulfilled the American College of Rheumatology (ACR) classification criteria for SLE (17), had onset of SLE before the age of 18 years, and underwent DXA scanning between January 2001 and July 2004 were eligible for the study.

The medical records of these patients were reviewed and information on the following parameters was extracted: ethnicity, sex, body mass index (BMI) at the time of DXA, age at diagnosis, age at the time of DXA, disease duration, corticosteroid requirement (duration of corticosteroid use, cumulative corticosteroid dose), requirement for and duration of other medications (methotrexate, cyclophosphamide, hydroxychloroquine, cyclosporine, mycophenolate mofetil, and azathioprine), clinical features (renal involvement including World Health Organization classification, central nervous system involvement, arthritis, serositis, and cutaneous involvement), osteoporosis complications (vertebral fractures), puberty status (pre- or postpubertal), quality of life (Childhood Health Assessment Questionnaire [18]), damage score (damage defined as a Systemic Lupus International Collaborating Clinics/ACR Damage Index [SDI] [19] of >0), disease activity (scores on the SLE Disease Activity Index [SLEDAI] [20] and European Consensus Lupus Activity Measure [ECLAM] [21]) at the time of DXA, and adjusted mean SLEDAI (22) during 6 months prior to the bone mineral density (BMD) assessment. Calcium and vitamin D supplementation at the time of DXA could not be accurately evaluated by chart review and therefore were not included in the analysis. Puberty status was determined either by using Tanner stage when available (puberty defined as Tanner stage >4) or by determining whether female patients had had their first menses prior to DXA. BMI was defined as weight/height2. Ethnicity was divided into 5 groupings: African, Asian, Caucasian, Native Canadian, and other (when ethnicity was mixed).

Determination of the average amount of physical activity was not possible by chart review. None of the patients were involved in any formal exercise program for osteoporosis.

The study was approved by the Research Ethics Board of the Hospital for Sick Children.

BMD measurement.

The first DXA scan performed on each patient was chosen for study. All BMD measurements were obtained with the same DXA instrument (Lunar Prodigy; GE Lunar, Madison, WI). BMD was measured in the lumbar spine (L2–L4) and in the femoral neck (mean value of the right and left hips was used). The lumbar spine BMD values were transformed into Z scores by comparing them with age- and sex-specific reference values for this equipment (23–25). The femoral neck BMD measurements were expressed as percentages since the sample size used for the pediatric reference values was not of sufficient magnitude to allow for use of a Z score. We defined osteopenia as a lumbar spine BMD Z score of <−1 and ≥−2.5, osteoporosis as a lumbar spine BMD Z score of <−2.5, and decreased hip BMD as a BMD of <80%.

Statistical analysis.

Descriptive statistics were used to assess the demographic variables. Pearson's correlation coefficient was used in univariate analysis to determine associations with osteopenia and osteoporosis. The chi-square test and Fisher's exact test were used as appropriate to examine associations between dichotomous variables. Variables identified as significant in the univariate analysis were entered into a logistic regression analysis. Statistical tests were performed using SAS, version 8.2 (SAS Institute, Cary, NC). P values less than 0.05 were considered significant.

RESULTS

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. REFERENCES

The cohort consisted of 64 patients (49 girls and 15 boys) who had a DXA scan performed (Table 1). DXA was performed a mean ± SD of 2.9 ± 2.8 years following diagnosis; in 5 patients, DXA was performed at presentation. Lumbar spine BMD was in the range of osteopenia in 24 patients (37.5%) and in the range of osteoporosis in 13 patients (20.3%). Twelve patients (18.75%) had a mean hip BMD of <80% (Table 1); the lumbar spine BMD Z score was <−1 in 10 of these patients and <−2.5 in 6. Clinical and laboratory characteristics of the total cohort and of BMD subgroups are shown in Table 1.

Table 1. Demographic and disease characteristics and medication use in the total cohort and in patients with lumbar spine osteopenia, patients with lumbar spine osteoporosis, and patients with decreased hip BMD*
CharacteristicTotal juvenile SLE cohort (n = 64)Patients with osteopenia (lumbar spine BMD <−1 and ≥−2.5) (n = 24)Patients with osteoporosis (lumbar spine BMD <−2.5) (n = 13)Patients with hip BMD <80% (n = 12)
  • *

    Except where indicated otherwise, values are the mean ± SD. BMD = bone mineral density; SLE = systemic lupus erythematosus; C-HAQ = Childhood Health Assessment Questionnaire; ECLAM = European Consensus Lupus Activity Measurement; SLEDAI = SLE Disease Activity Index; SDI = Systemic Lupus International Collaborating Clinics/American College of Rheumatology Damage Index; CNS = central nervous system; NSAID = nonsteroidal antiinflammatory drug.

  • Ethnicity data were not available for some patients.

Female, no. (%)49 (76.6)17 (70.8)8 (61.5)5 (41.7)
Age, years14.3 ± 3.315.8 ± 2.816.6 ± 2.114.74 ± 3.68
Ethnicity, no. (%)    
 Caucasian14 (26.9)6 (27.3)4 (33.3)2 (18.2)
 Asian22 (42.3)11 (50)5 (41.7)7 (63.6)
 Other16 (30.8)5 (22.7)3 (25)2 (18.2)
Postpubertal, no. (%)40 (62.5)16 (66.7)9 (69.2)4 (33.3)
Body mass index, kg/m222.2 ± 4.822.3 ± 5.023.2 ± 4.722.3 ± 5.63
Age at disease onset, years11.4 ± 3.411.4 ± 3.610.8 ± 4.210.05 ± 3.48
Disease duration, years2.9 ± 2.84.4 ± 3.55.8 ± 3.74.69 ± 3.43
C-HAQ >0, no. (%)13 (36.1)4 (26.7)1 (12.5)1 (12.5)
ECLAM2.6 ± 2.33.0 ± 2.42.3 ± 1.22.78 ± 2.73
SLEDAI4.6 ± 4.94.9 ± 5.44.1 ± 3.44.33 ± 5.52
Adjusted mean SLEDAI3.5 ± 3.14.2 ± 3.65.2 ± 3.83.09 ± 3.23
SDI >0, no. (%)10 (16.9)7 (33.3)5 (50)3 (33.3)
Clinical features, no. (%)    
 Skin disease50 (78.1)17 (70.8)8 (61.5)10 (83.3)
 Serositis15 (23.4)7 (29.2)5 (38.5)6 (50)
 CNS involvement12 (18.7)6 (25)3 (23.1)2 (16.7)
 Arthritis42 (65.6)13 (54.2)7 (53.8)5 (41.7)
 Lupus nephritis38 (59.4)19 (79.2)12 (92.3)10 (83.3)
 Class III–IV nephritis32 (50)15 (62.5)10 (76.9)9 (75)
Medications used, no. (%)    
 NSAID23 (35.9)9 (37.5)5 (38.5)5 (41.7)
 Methotrexate7 (10.9)3 (12.5)1 (7.7)2 (16.7)
 Azathioprine34 (53.1)18 (75)11 (84.6)7 (58.3)
 Mycofenolate mofetil7 (10.9)4 (16.7)4 (30.8)3 (25)
 Cyclophosphamide14 (21.9)9 (37.5)7 (53.8)5 (41.7)
 Cyclosporine1 (1.6)1 (4.2)0 (0)0 (0)
 Hydroxychloroquine45 (70.3)17 (70.8)8 (61.5)6 (50)
Duration of medication use, days    
 Methotrexate634.9 ± 585.51,101.3 ± 607.21,330.0710.5 ± 876.1
 Azathioprine893.4 ± 815.91,153.0 ± 937.31,441.1 ± 1,032.62,008.8 ± 804.2
 Mycophenolate mofetil409.3 ± 413.6380.2 ± 445.1380.2 ± 445.1171 ± 185.8
 Hydroxychloroquine982.1 ± 90.11,360.8 ± 1,117.51,829.6 ± 1,238.91,546.3 ± 1,352.0
Corticosteroid    
 Ever used, no. (%)62 (96.9)24 (100)13 (100)12 (100)
 Current user, no. (%)62 (96.9)24 (100)13 (100)12 (100)
 Prednisone, no. (%)62 (96.9)24 (100)13 (100)12 (100)
 Methylprednisolone, no. (%)13 (20.3)7 (29.2)4 (30.8)2 (16.7)
 Cumulative dose, gm/kg319.5 ± 342.1506.2 ± 413.6626.9 ± 381.5560.9 ± 417.5
 Duration of use, years2.7 ± 2.63.9 ± 3.25.1 ± 3.44.1 ± 2.9

Univariate analysis.

The clinical, laboratory, and medication data were entered into univariate analyses for osteopenia at the lumbar spine, osteoporosis at the lumbar spine, and mean hip BMD <80%. This analysis revealed that osteopenia was associated with the age of the patient at the time of DXA, disease duration, duration of corticosteroid therapy, cumulative corticosteroid dose, azathioprine use, cyclophosphamide use, history of nephritis, and presence of damage (SDI >0) (Table 2). Osteoporosis was associated with the age of the patient at the time of DXA, disease duration, duration of corticosteroid therapy, cumulative corticosteroid dose, azathioprine use, mycophenolate mofetil use, cyclophosphamide use, history of nephritis, renal biopsy showing class III–IV lupus nephritis, and presence of damage (Table 2). Mean hip BMD <80% was associated with disease duration, duration of corticosteroid use, and cumulative corticosteroid dose (Table 2). Of note, we did not find that any statistically significant associations between scores on any of the measures of disease activity (SLEDAI, ECLAM, adjusted mean SLEDAI) and either osteopenia or osteoporosis at the lumbar spine or hip BMD <80%.

Table 2. Univariate analysis of clinical variables and medications*
VariableP
Lumbar spine BMD <−1 and ≥−2.5Lumbar spine BMD <−2.5Hip BMD <80%
  • *

    All variables listed in Table 1 were tested; only the statistically significant variables are shown here. BMD = bone mineral density; DXA = dual x-ray absorptiometry; SDI = Systemic Lupus International Collaborating Clinics/American College of Rheumatology Damage Index.

Age at time of DXA0.00960.00950.6170
Disease duration0.0030.00090.0249
Duration of corticosteroid use0.00490.00220.0497
Cumulative corticosteroid dose0.00260.00170.0125
Azathioprine use0.00850.02020.6888
Mycophenolate mofetil use0.26660.02040.1012
Cyclophosphamide use0.02440.00380.0756
Lupus nephritis0.01570.02330.0770
Class III or IV lupus nephritis0.12450.03880.0657
SDI >00.02020.00590.1691

Multivariate analysis.

Values that were statistically significantly associated with lumbar spine osteopenia, lumbar spine osteoporosis, and/or decreased hip BMD were entered separately into regression analyses. Our initial analysis (Pearson's correlation test) revealed that duration of corticosteroid use, disease duration, and age at the time of DXA were highly correlated, making the model unstable (Table 3). Disease duration and duration of corticosteroid use were particularly interrelated, with a Pearson correlation coefficient of 0.97 (P <0.0001) (Table 3). As a result of this finding, 3 separate analyses were performed using 2 of these 3 variables at a time, for lumbar spine osteopenia, lumbar spine osteoporosis, and mean hip BMD <80%. The best models for each cutoff are shown in Table 4. Cumulative corticosteroid dose was significantly associated with lumbar spine osteopenia, while disease duration best predicted both lumbar spine osteoporosis and mean hip BMD <80%. The association between nephritis and lumbar spine osteoporosis approached statistical significance (P = 0.0541) (Table 4).

Table 3. Pearson's correlation coefficients between 3 interrelated variables assessed in pairs
VariableCorrelation coefficientP
  • *

    DXA = dual x-ray absorptiometry.

Age at time of DXA/disease duration*0.381190.0019
Age at time of DXA/duration of corticosteroid use0.383780.0017
Disease duration/duration of corticosteroid use0.97341<0.0001
Table 4. Multivariate analysis results*
 Variable (P)
  • *

    Three separate multivariate analyses using 2 of 3 interrelated variables at a time were performed. All variables that were statistically significant in univariate analysis were entered into the regression analysis. BMD = bone mineral density.

Correlation with lumbar spine BMD <−1 and ≥−2.5Cumulative corticosteroid dose (0.0026)
Correlation with lumbar spine BMD <−2.5Disease duration (0.0028), lupus nephritis (0.0541)
Correlation with hip BMD <80%Disease duration (0.0249)

Further analysis of the statistically significant parameters confirmed the validity of the models, since odds ratio point estimates were >1 with confidence intervals excluding 1. The results of this analysis confirmed that cumulative corticosteroid dose was significantly associated with lumbar spine osteopenia and that disease duration best predicted both lumbar spine osteoporosis and mean hip BMD <80% (Table 5). This analysis also showed that the association between lupus nephritis and lumbar spine osteoporosis was not statistically significant but approached significance.

Table 5. Results of the final models (multivariate analysis)*
 Variable (OR point estimate [Wald's 95% CI])
  • *

    The statistical significance of the models identified in the multivariate analysis was further analyzed, and the validity of the models was confirmed. OR = odds ratio; 95% CI = 95% confidence interval; BMD = bone mineral density.

Correlation with lumbar spine BMD <−1 and ≥−2.5Cumulative corticosteroid dose (1.003 [1.001–1.005])
Correlation with lumbar spine BMD <−2.5Disease duration (1.599 [1.175–2.176]), lupus nephritis (8.788 [0.962–80.235])
Correlation with hip BMD <80%Disease duration (1.290 [1.033–1.611])

DISCUSSION

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. REFERENCES

Osteoporosis is a systemic skeletal disorder characterized by low bone mass and microarchitectural deterioration of bone tissue, leading to increased bone fragility and fracture (26). Although osteoporosis is a well-known complication of adult-onset SLE, few studies have been performed in juvenile SLE (2, 14–16). In contrast to studies of juvenile dermatomyositis (11) and juvenile idiopathic arthritis (9, 12, 13, 27, 28), which have shown decreased BMD in these patients, only 2 of 4 studies of juvenile SLE demonstrated decreased BMD (14, 16). In this study, we found that osteopenia and osteoporosis at the lumbar spine and low hip BMD were common in patients with juvenile SLE, occurring in 38%, 20%, and 19% of patients, respectively. The major difference between our study and previous investigations in juvenile SLE is that the sample size of pediatric patients in our study was larger than those in the earlier studies in which lower BMD in patients with juvenile SLE was also demonstrated (14, 16). Consistent with our results were the findings of 2 studies of patients who were >19 years old at the time of study but whose SLE had begun during childhood, both of which demonstrated decreased BMD (2, 14). While the frequency of osteopenia was comparable (14), the frequency of osteoporosis was higher in our study, in which the population was exclusively pediatric.

Because the results of previous work in adults with SLE have suggested that both chronic inflammation and corticosteroid therapy are associated with decreased BMD, we examined the contributions of these factors to osteopenia and osteoporosis in order to better understand the pathogenesis of decreased BMD in patients with juvenile SLE. We did not find that the cumulative dose of corticosteroids or duration of corticosteroid use was an important predictor of lumbar spine osteoporosis or hip BMD <80%. However, cumulative corticosteroid dose was associated with lumbar spine osteopenia by multivariate analysis, although the low odds ratio suggested that the contribution was minor.

Two previous studies in juvenile SLE showed that BMD was correlated with cumulative corticosteroid dose, but the study populations included adults with childhood-onset SLE, as well as children (2, 14). Consistent with our findings are the results of studies in adult premenopausal patients with SLE, in which decreased BMD was not found to be correlated with corticosteroid use, total corticosteroid dose, or duration of corticosteroid use (29–37). Investigations of pediatric patients with nephrotic syndrome have yielded conflicting results regarding the role of corticosteroids in the development of osteoporosis (38, 39). One reason for the disparate findings regarding the role of corticosteroids in osteoporosis might relate to the fact that prolonged high-dose steroid therapy may lead to increases in weight and BMI, which may be associated with secondary improvement of spinal bone mineral content. Further studies are needed to confirm our findings of the lack of association of osteoporosis with corticosteroid use in juvenile SLE.

Studies in adults with SLE have suggested that the decreased BMD is related to disease damage and disease duration, rather than to measures of disease activity or severity (30, 32–34, 36). In the present study, univariate analysis showed multiple measures of disease activity, specific disease manifestations, and therapies to be associated with the 3 measures of abnormally low BMD. However, multivariate analysis revealed that only disease duration was associated with lumbar spine osteoporosis and decreased hip BMD. The only measure of disease activity that approached statistical significance was the association of lupus nephritis as a predictor of lumbar spine osteoporosis. The role of this variable must be analyzed carefully and may require further study. Interestingly, whether the nephritis was active or inactive at the time of BMD measurement did not affect this finding. Previous studies of patients with juvenile SLE failed to show a relationship between BMD and measures of disease activity, severity, or damage (2, 14–16).

Consistent with studies in adults with SLE (33, 34), we did not find that scores on global measures of disease activity, including the SLEDAI, ECLAM, and adjusted mean SLEDAI, were significantly associated with low BMD, despite evidence that these indices are good measures of disease activity in juvenile SLE (40–42). Although we found that many surrogate measures of disease severity, including use of immunosuppressive agents and disease damage score (as measured by SDI), were significantly associated with BMD by univariate analysis, none of these measures was significantly associated with BMD by multivariate analysis. Our finding of a lack of association with measures of disease damage in multivariate analysis contrasts with the results of studies in adult SLE (30, 32–34, 36). Our results suggest that factors other than disease damage or severity may be more important in determining BMD in juvenile SLE. However, these findings might also be explained by a lack of power in our analysis, given the sample size.

To date, there has not been a standard way of describing BMD results or defining osteopenia and osteoporosis in pediatric patients with rheumatic diseases (2, 16). Definitions of osteopenia and osteoporosis in adults (43) have been validated by epidemiologic studies correlating each standard deviation decrease in BMD with an increase of fracture risk, in postmenopausal women (44–46). The validity of using Z scores and definitions of adult osteopenia and osteoporosis in pediatric patients has been demonstrated in a study of patients with juvenile dermatomyositis (47). In the present work we report lumbar spine BMD as Z scores since these values were available for this site in our patients. Hip BMD is reported as percentages, since validated Z scores were not available. At our institution, hip BMD of 80–90% is considered a mild decrease (equivalent to a Z score of approximately −1 to −2 in adults), 70–80% a moderate decrease, and <70% a marked decrease. Therefore, we chose 80% (0.8) as a cutoff for defining abnormal hip BMD in the present study. We suggest that in general, the definitions of osteopenia and osteoporosis used in adults, i.e., scores of <−1 and <−2.5, respectively, should be used in patients with pediatric rheumatic diseases. However, a large collaborative study is needed to confirm that these cutoffs define increased risk of fracture in children, as has been done in postmenopausal women.

A limitation of our study was the use of BMD, which does not take into account structural properties of bone (5, 25). It would have been of benefit to have data regarding microarchitecture and mineralization of the bones, but we could not justify performing bone biopsy, nor did we have access to quantitative measures obtained by computed tomography or magnetic resonance imaging. However, DXA data can be readily extrapolated to clinical practice. Another limitation of our study was the lack of data on calcium and vitamin D intake, due to the retrospective design of the study. Although calcium supplementation has been shown to be of benefit in maintaining normal BMD increase in healthy prepubertal children (7), studies of children with rheumatic diseases have yielded conflicting results (14, 16, 48–50). Similarly, we did not evaluate the effect of activity on BMD, due to the retrospective study design. Although regular exercise is known to increase BMD in healthy children (51–53), studies in patients with pediatric rheumatic diseases have had conflicting results (16, 28). Prospective studies are needed to determine the effects of calcium and vitamin D intake and exercise on BMD and fracture risk in juvenile SLE.

Our study demonstrated rates of osteopenia and osteoporosis at the lumbar spine of 38% and 20%, respectively, and a 19% rate of decreased hip BMD. These findings are similar to those in previous investigations of adult SLE (29–37). We suggest that, pending future studies examining the association of fracture risk and BMD in pediatric patients, the definitions of osteopenia and osteoporosis used in adults, i.e., −1 and −2.5, respectively, should be used. Disease duration and possibly lupus nephritis (not necessarily active at the time of DXA) were identified as predictors of low BMD, suggesting that patients with juvenile SLE, and in particular those with nephritis, are at risk of osteoporosis and possibly fracture. However, since the long-term safety of antiresorptive agents has not been established in pediatric patients, we do not advocate the routine use of these agents to prevent osteoporosis in children with longstanding SLE. Long-term studies are needed to determine the morbidity associated with osteoporosis in juvenile SLE and the role of prophylactic therapy to prevent this complication.

AUTHOR CONTRIBUTIONS

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. REFERENCES

Dr. Silverman had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study design. Compeyrot-Lacassagne, Gilday, Silverman.

Acquisition of data. Compeyrot-Lacassagne, Tyrrell, Gilday, Silverman.

Analysis and interpretation of data. Compeyrot-Lacassagne, Tyrrell, Atenafu, Doria, Stephens, Silverman.

Manuscript preparation. Compeyrot-Lacassagne, Tyrrell, Atenafu, Stephens, Silverman.

Statistical analysis. Atenafu, Stephens.

REFERENCES

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. REFERENCES