Skeletal Outcomes in Children and Young Adults with Glomerular Disease

Significance Statement

Children and young adults with glomerular disease have unique risk factors for compromised bone health, but data on skeletal complications are lacking. We leveraged the PEDSnet pediatric health system population of more than 6.5 million children to compare incidence rates of adverse skeletal outcomes in patients with glomerular disease to a general pediatric reference cohort. Children and young adults with glomerular disease had greater risk of vertebral and hip/femur fractures, avascular necrosis/osteonecrosis, and slipped capital femoral epiphysis than those in the reference cohort. For fractures at any body site, girls with glomerular disease were at increased risk compared with peers in the general pediatric population, an effect that CKD does not appear to drive entirely.

Visual Abstract

Abstract

Background

Children with glomerular disease have unique risk factors for compromised bone health. Studies addressing skeletal complications in this population are lacking.

Methods

This retrospective cohort study utilized data from PEDSnet, a national network of pediatric health systems with standardized electronic health record data for more than 6.5 million patients from 2009 to 2021. Incidence rates (per 10,000 person-years) of fracture, slipped capital femoral epiphysis (SCFE), and avascular necrosis/osteonecrosis (AVN) in 4598 children and young adults with glomerular disease were compared with those among 553,624 general pediatric patients using Poisson regression analysis. The glomerular disease cohort was identified using a published computable phenotype. Inclusion criteria for the general pediatric cohort were two or more primary care visits 1 year or more apart between 1 and 21 years of age, one visit or more every 18 months if followed >3 years, and no chronic progressive conditions defined by the Pediatric Medical Complexity Algorithm. Fracture, SCFE, and AVN were identified using SNOMED-CT diagnosis codes; fracture required an associated x-ray or splinting/casting procedure within 48 hours.

Results

We found a higher risk of fracture for the glomerular disease cohort compared with the general pediatric cohort in girls only (incidence rate ratio [IRR], 1.6; 95% CI, 1.3 to 1.9). Hip/femur and vertebral fracture risk were increased in the glomerular disease cohort: adjusted IRR was 2.2 (95% CI, 1.3 to 3.7) and 5 (95% CI, 3.2 to 7.6), respectively. For SCFE, the adjusted IRR was 3.4 (95% CI, 1.9 to 5.9). For AVN, the adjusted IRR was 56.2 (95% CI, 40.7 to 77.5).

Conclusions

Children and young adults with glomerular disease have significantly higher burden of skeletal complications than the general pediatric population.

Numerous studies have demonstrated increased fracture rates and associated morbidity and mortality in adults on dialysis^1–6 and adults with moderate to advanced (stage 3–5) CKD.^7–12 Pediatric studies have also found a high burden of fracture in children with CKD.^13,14 None of these studies evaluated the effect of underlying glomerular disease. With 90% of peak bone mass established by 18 years of age, threats to healthy bone development in childhood have a critical effect on lifelong skeletal health. Glomerular disorders are frequent causes of CKD in children and young adults, although the majority of those with idiopathic nephrotic syndrome have early-stage CKD and normal kidney function. Even in the setting of normal kidney function, however, individuals with glomerular disease have unique and potentially modifiable risk factors for skeletal morbidity, such as disordered vitamin D homeostasis, exposure to medications with adverse skeletal effects such as corticosteroids, and chronic systemic inflammation. Although prior studies have demonstrated reduced bone mineral density in children with nephrotic syndrome^15–25 and in young adults with a history of nephrotic syndrome in childhood,^26,27 studies addressing fractures and other skeletal complications among these populations are lacking.

Glomerular diseases are rare, and studies of clinical outcomes in children with glomerular disease have been largely limited to case series, with a paucity of population-based data.¹⁸ This study was conducted within PEDSnet, a multi-institutional national clinical research network, and leveraged a computable phenotype algorithm that was demonstrated to identify children with glomerular disease reliably,²⁸ allowing for both complete capture of this population and comparison to a reference cohort of general pediatric patients within the network.

The main objective of this retrospective cohort study was to examine the incidence rates of skeletal outcomes, including fracture, avascular necrosis/osteonecrosis (AVN), and slipped capital femoral epiphysis (SCFE), in children and young adults with glomerular disease compared with those among general pediatric patients. In secondary analyses, we examined rates of skeletal outcomes in the subcohort of children with nephrotic conditions, and how moderate to advanced CKD affected rates of skeletal outcomes in glomerular disease. Finally, we investigated the effect of corticosteroid exposure on the risk of skeletal outcomes within the cohort of children with glomerular disease.

Methods

Study Setting

PEDSnet (pedsnet.org) is a multi-institutional clinical research network that aggregates electronic health record (EHR) data from several of the nation’s large children’s health care organizations (Institutional Review Board [IRB] protocol #14–011242). Seven institutions participated in this study: Children’s Hospital of Philadelphia (CHOP), Cincinnati Children’s Hospital Medical Center (CCHMC), Children’s Hospital Colorado, Nationwide Children’s Hospital, Nemours Children’s Health System (a Delaware and Florida health system), Seattle Children’s Hospital, and St. Louis Children’s Hospital/Washington University.²⁹ PEDSnet standardizes EHR data across institutions to the PEDSnet common data model, which is based on the Observational Medical Outcomes Partnership common data model version 5.³⁰ PEDSnet has accrued data for nearly seven million children seen since 2009 from inpatient and outpatient clinical settings. The PEDSnet database includes data from patients who reside in all 50 states; the states contributing the largest volume of data are Colorado, Delaware, Florida, Illinois, Indiana, Kentucky, Missouri, New Jersey, Ohio, Pennsylvania, and Washington. This analysis was reviewed by the IRB at CHOP and CCHMC because, at the time of analysis, CCHMC housed its PEDSnet data separately, whereas data from the other centers were centrally stored at CHOP. This study was determined to meet criteria for exemption by the CCHMC IRB and was determined not to meet the criteria for human subjects research by the CHOP IRB, which was accepted by all other PEDSnet sites.

Cohort Definitions

Figure 1 outlines the criteria for derivation of the two central cohorts under study: children with glomerular disease and general pediatric patients receiving primary care. Among the glomerular disease cohort, a subcohort of children with nephrotic conditions was also evaluated.

Figure 1.

Derivation of study cohorts: children with glomerular disease and general pediatric patients receiving primary care.

The cohort of 4598 children with glomerular disease was defined using a computable phenotype algorithm, which was previously developed and evaluated in PEDSnet.²⁸ This algorithm involves two or more glomerular disease diagnosis codes on different dates or one or more glomerular disease diagnosis code(s) and a native kidney biopsy procedure. An associated computational phenotype was used to identify the subcohort of 2238 children and young adults with nephrotic conditions, including idiopathic nephrotic syndrome (without a biopsy), minimal change nephropathy, focal segmental glomerulosclerosis, and membranous nephropathy. Patients were required to have at least one encounter with a nephrology provider or at a nephrology clinic. Follow-up time for this cohort began at the first encounter with a glomerular disease diagnosis and ended at the most recent “in-person” clinical encounter (inpatient, outpatient, or emergency department visit).

To develop a cohort of children and young adults representative of the general pediatric population, we selected patients receiving primary care, excluding those for whom there was evidence in the EHR of severe chronic disease. The general pediatric cohort was defined as follows. First, we restricted the cohort to the three health systems within PEDSnet with large regional primary care networks (CHOP, Nemours, and Nationwide). Second, we required evidence of ongoing primary care, defined as at least two outpatient general pediatric or family medicine encounters and at least one such encounter every 18 months if followed >3 years (approximating recommendations for frequency of well-child visits). Third, we applied the taxonomy from the Pediatric Medical Complexity Algorithm (PMCA)³¹ to identify and exclude patients with a diagnosis associated with deteriorating health or reduced life expectancy, labeled as “progressive” in the PMCA. In addition, patients meeting the criteria for CKD (defined below) were excluded. Follow-up time for this cohort began at the first outpatient general pediatric encounter and ended at the last outpatient general pediatric encounter.

We applied a set of three additional requirements to both the glomerular disease and general pediatric cohorts. First, we required that all patients have at least 1 year of follow-up, defined as two outpatient visits separated by at least 12 months. The motivation for this criterion was to avoid inclusion of patients who interact with the network for episodic care as the amount of person-time contributed per outcome event would be small and inflate incidence rates. Second, we excluded patients who had more orthopedic-related visits than all other visit types combined. The motivation for this exclusion was to remove patients from the general pediatric cohort with recurrent orthopedic issues because they are not representative of the general population; further, for both cohorts, having more orthopedic than all other visit types could indicate that nonorthopedic care (for kidney disease or primary care) was being received largely outside the network. Third, we implemented a cancer diagnosis exclusion because of established skeletal morbidity in survivors of childhood cancer.³²

Assessment of Skeletal Outcomes

We developed diagnosis code sets using the SNOMED terminology for bone fracture, AVN, and SCFE. All code sets used are available at https://github.com/PEDSnet/skeletal_outcomes. For bone fracture, given a concern that some diagnoses in the EHR would represent prevalent rather than incident fractures, we required evidence of an accompanying radiologic, splinting, or casting procedure within 48 hours of the fracture diagnosis. Due to the challenge of distinguishing recurrent skeletal outcome events from a single skeletal outcome event receiving ongoing treatment within a period of several months within the EHR, we took an annualized approach to fracture within each patient year of age in order to avoid inflation of incidence rates due to capturing encounters for follow-up care of the skeletal outcome. For each patient, if their follow-up period included any time within a given year of age, they contributed person-time and a determination of whether they experienced at least one fracture within that year of age. When follow-up began or ended within a year of age, patients contributed the number of days of follow-up within that year of age. We excluded person-time below the age of 2 years for several reasons. Fractures occurring before 18 months, the usual age of walking, have different risk factors.^33,34 Measures such as body mass index and eGFR, which we included in our analyses, are not appropriately applied before 2 years of age. Finally, as glomerular disease diagnoses are rare before 2 years of age, sample sizes were small for the first and second year of life.

Primary Analyses

Standard descriptive statistics were used to summarize cohort characteristics, utilizing medians and interquartile ranges for continuous variables and counts and frequencies for categorical variables. Unadjusted incidence was calculated as the number of events per 10,000 person-years. Overall incidence (incidence across the age span of 2–21 years) was calculated as the total number of annualized events divided by the total follow-up for the cohort across the age span. For incidence within a given year of age, the number of annualized events within the year of age was divided by total person-time within that same year of age.

We applied Poisson regression to model the aggregated number of skeletal outcome events within each year of age group, stratified by covariates, using total person-time within each year of age as the offset. It is well established that rates of skeletal outcomes vary across age and sex.³⁵ We therefore adjusted for year of age, sex, and the interaction between age and sex across the analyses, and also performed sex-stratified analyses. To capture any potential nonlinear relationship between the skeletal outcome events and age, we included polynomial terms of age (i.e., age, age squared, and age cubed) in the regression models. We also adjusted for race and obesity because obesity is a known risk factor for adverse skeletal outcomes,^36–38 racial differences in bone outcomes have been reported in healthy children and children with CKD,³⁹ and the distribution of race and obesity differed across cohorts. Obesity was operationalized as a non-time-updated binary variable corresponding to whether a patient had three or more body mass index measures at or above the 95th percentile, separated by at least 6 months, between 2 years of age and the end of their follow-up. Race (from the EHR) was categorized as: Asian American, Black or African American, other (American Indian or Alaska Native, Native Hawaiian or Other Pacific Islander, Multiple, other), unknown, or White. Interactions were retained in the model only when they were significant by the likelihood ratio test. We evaluated certain site-specific fractures (vertebra and hip/femur), for which we did not perform sex-stratified analyses due to the lower incidence of these outcomes. Incidence rate ratios (IRR) with 95% confidence intervals (CI) were reported between cohorts.

Secondary Analyses

CKD was evaluated as a time-updated exposure in the glomerular cohort where person-time before meeting the CKD criteria was considered unexposed and person-time thereafter exposed, if applicable. A patient was defined as meeting the criteria for CKD when there was evidence in the EHR of one or more of the following: a dialysis procedure, a kidney transplant, or two eGFR measures of <60 ml/min per 1.73 m² separated by 90 days (without a subsequent eGFR>90 ml/min per 1.73 m², unless this eGFR followed a kidney transplant). The “bedside” estimating equation developed by the Chronic Kidney Disease in Children study was used to calculate eGFR⁴⁰ on the basis of serum creatinine and the closest associated height measurement within 60 days.

Cox proportional hazard models were applied to investigate whether corticosteroid exposure, operationalized as a fixed covariate of two corticosteroid prescriptions separated by at least 90 days, affected time to first fracture and time to AVN within the glomerular disease cohort, adjusting for sex, age, race, and obesity. In these analyses, patients were right censored at end of follow-up or first date at which CKD criteria were met. The proportional hazard assumption was examined using the Pearson product-moment correlation between the scaled Schoenfeld residuals and log-transformed time for each covariate. Supplemental Table 1 provides a summary of all analyses performed.

Results

Descriptive Characteristics

Descriptive characteristics of each cohort are consistent with clinical expectations (Table 1). For proteinuria, in the glomerular disease cohort and nephrotic subcohort, >80% of patients had at least one positive urine protein measurement compared with 6.9% in the general pediatric cohort. About half of the glomerular cohort had a hypertension diagnosis, and 42.5% had a kidney biopsy. Systemic corticosteroid exposure was seen in 62.7%, 77.4%, and 23.5% of the glomerular, nephrotic, and general pediatric cohorts, respectively. For the nonsteroid immunosuppressants included in Table 1, rates were consistently close to zero in the general pediatric cohort. More than 50% of the glomerular disease cohort and nephrotic subcohort had at least one hospitalization compared with 7.1% of the general pediatric cohort. Obesity was identified in 24.8% of the glomerular disease cohort and 27.2% of the nephrotic subcohort compared with 14.7% in the general pediatric cohort.

Table 1.

Population characteristics

	Cohort
	Glomerular Disease	Nephrotic Subcohort	General Pediatric
N	4598	2238	553,624
Age at index in years	9.83 (5.78, 14.7)	7.59 (4.2, 13.4)	2.9 (2.05, 7.24)
Sex
Boys	2514 (54.7)	1233 (55.1)	279,821 (50.5)
Girls	2084 (45.3)	1005 (44.9)	273,803 (49.5)
Race
Black or African American	860 (18.7)	507 (22.7)	175,608 (31.7)
Other	698 (15.2)	359 (16)	33,341 (6.02)
Unknown	264 (5.74)	149 (6.66)	57,045 (10.3)
White	2582 (56.2)	1112 (49.7)	265,874 (48)
Asian American	194 (4.22)	111 (4.96)	21,756 (3.93)
Ethnicity
Hispanic or Latinx	718 (15.6)	346 (15.5)	52,900 (9.56)
Other	36 (0.783)	16 (0.715)	1866 (0.337)
Not Hispanic or Latinx	3673 (79.9)	1790 (80)	453,926 (82)
Unknown	171 (3.72)	86 (3.84)	44,932 (8.12)
Years of follow-up duration	4.05 (2.04, 7.17)	4.47 (2.32, 7.87)	6.55 (3.29, 10.6)
≥1 nephrology visit	4438 (96.5)	2177 (97.3)	6157 (1.11)
Nephrology visits per person-year	2.71 (1.09, 5.3)	3.01 (1.43, 5.74)	0 (0, 0)
≥1 general pediatric visit	1630 (35.5)	809 (36.1)	553,624 (100)
General pediatric visits per person-year	0 (0, 0.715)	0 (0, 0.755)	2.65 (1.77, 3.84)
≥1 25-hydroxyvitamin D measurement	618 (13.4)	266 (11.9)	4104 (0.741)
Height measured	4564 (99.3)	2218 (99.1)	551,550 (99.6)
Z-scored height at index	−0.079 (–0.883, 0.663)	−0.117 (–0.913, 0.575)	0.262 (–0.767, 0.934)
≥3 BMI >95th percentile	1139 (24.8)	609 (27.2)	81,370 (14.7)
BMI measured	4566 (99.3)	2218 (99.1)	551,263 (99.6)
Z-scored BMI at index	0.862 (–0.00043, 1.67)	1.11 (0.252, 1.86)	0.232 (–0.522, 0.982)
≥1 serum creatinine measurement	4380 (95.3)	2117 (94.6)	104,448 (18.9)
Serum creatinine measurements per person-year	3.35 (1.09, 9.71)	2.71 (0.912, 8.04)	0 (0, 0)
Minimum eGFR (ml/min per 1.73 m2) across follow-up, right censored at kidney transplant or dialysis
<15	609 (13.2)	247 (11)	222 (0.0401)
15–<30	294 (6.39)	124 (5.54)	45 (0.00813)
30–<60	579 (12.6)	289 (12.9)	626 (0.113)
60–<90	1095 (23.8)	492 (22)	19,889 (3.59)
≥90	1422 (30.9)	802 (35.8)	59,515 (10.8)
Unavailable	599 (13)	284 (12.7)	473,327 (85.5)
≥1 urine protein measurement	3896 (94.4)	1904 (94.4)	125,131 (22.6)
Urine protein measurements per person-year	2.99 (1.11, 6.63)	3.1 (1.14, 6.78)	0 (0, 0)
≥1 positive urine protein measurement	3467 (84)	1743 (86.4)	38,291 (6.92)
≥1 hypertension diagnosis	2323 (50.5)	1029 (46)	7536 (1.36)
≥1 hospitalization	2524 (54.9)	1327 (59.3)	39,176 (7.08)
Hospitalizations per person-year	0.146 (0, 0.685)	0.193 (0, 0.722)	0 (0, 0)
≥1 kidney biopsy procedure	1953 (42.5)	902 (40.3)	11 (0.00199)
ACE inhibitors	1902 (41.4)	866 (38.7)	659 (0.119)
Angiotensin II receptor blockers	616 (13.4)	316 (14.1)	140 (0.0253)
Aldosterone receptor blockers	114 (2.48)	66 (2.95)	999 (0.18)
Systemic corticosteroids	2883 (62.7)	1732 (77.4)	129,886 (23.5)
Mycophenolate	1182 (25.7)	573 (25.6)	21 (0.00379)
Calcineurin inhibitors	974 (21.2)	729 (32.6)	16 (0.00289)
Cyclophosphamide	324 (7.05)	150 (6.7)	<11 (<0.00199)
Azathioprine	263 (5.72)	54 (2.41)	91 (0.0164)
Rituximab	524 (11.4)	325 (14.5)	<11 (<0.00199)
Eculizumab	24 (0.522)	<11 (<0.492)	0 (0)
Corticotropin	<11 (<0.239)	<11(<0.492)	0 (0)
Abatacept	15 (0.326)	<11 (<0.492)	11 (0.00199)

All continuous data as median (25th percentile, 75th percentile); all categorical data as n (%). Urine protein measurements exclude one site, for which these measurements are unavailable. All medication and biopsy data are pre kidney transplant, if applicable. BMI, body mass index.

Incidence of Fracture in the General Pediatric Cohort

In the general pediatric cohort, the peak fracture incidence rate for boys was 270 per 10,000 person-years at age 13, whereas for girls, it was 138 per 10,000 person-years at age 11 (Figure 2). In the model for the general pediatric cohort (Supplemental Table 2), the linear, quadratic, and cubic components of age were significant predictors of fracture incidence, as expected given the curvilinear relationship between age and fracture risk, whereby risk is heightened around the time of peak linear growth during puberty. Obesity was associated with an increased incidence of fracture (IRR=1.2; 95% CI, 1.1 to 1.25).

Figure 2.

Age-specific fracture incidence by sex in the general pediatric cohort.

Incidence of Fracture in the Glomerular Disease Cohort Compared with the General Pediatric Cohort

The model comparing the glomerular disease cohort to the general pediatric cohort indicated an interaction between cohort and sex (Supplemental Table 3), whereby the effect of the cohort was strengthened in girls and attenuated in boys. Sex-stratified analyses demonstrated that among girls, the difference in fracture incidence between the cohorts changed across age (P=0.04) with a greater magnitude of difference before 10 years of age (Figure 3, Table 2). To understand whether there was an overall main effect of cohort among girls, we examined a model that excluded this interaction between cohort and age. In this model, girls with glomerular disease were 1.6 (95% CI, 1.3 to 1.9) times more likely to experience a fracture than girls in the general pediatric cohort (Table 3). The difference between the glomerular disease and general pediatric cohorts did not reach significance in the sex-stratified analyses for boys (Table 4). Obesity remained associated with increased incidence of fracture in sex-stratified analyses.

Figure 3.

Age-specific fracture incidence by sex in the glomerular disease cohort versus the general pediatric cohort.

Table 2.

Poisson regression model for incident fracture among girls in the glomerular disease and general pediatric cohorts

Variable	Incidence Rate Ratio	95% CI	P Value
Cohort: glomerular disease	0.986	0.182 to 5.336	0.99
Age (yr)	1.129	1.003 to 1.27	0.04
Age squared	1.005	0.992 to 1.018	0.42
Age cubed	0.999	0.999 to 1	0.001
Race
White	1	—	—
Black or African American	0.975	0.907 to 1.048	0.49
Other	1.339	1.221 to 1.467	<0.001
Asian American	0.745	0.68 to 0.817	<0.001
Unknown	0.934	0.81 to 1.076	0.35
Obese	1.31	1.215 to 1.413	<0.001
Interaction between cohort and age	1.356	0.792 to 2.321	0.27
Interaction between cohort and age squared	0.957	0.91 to 1.007	0.09
Interaction between cohort and age cubed	1.002	1 to 1.003	0.02

Table 3.

Poisson regression model for incident fracture among girls in the glomerular disease and general pediatric cohorts excluding a cohort and age interaction

Variable	Incidence Rate Ratio	95% CI	P Value
Cohort: glomerular disease	1.594	1.324 to 1.918	<0.001
Age (yr)	1.145	1.017 to 1.288	0.03
Age squared	1.003	0.991 to 1.016	0.59
Age cubed	0.999	0.999 to 1	0.003
Race
White	1	—	—
Black or African American	0.975	0.906 to 1.049	0.5
Other	1.339	1.22 to 1.468	<0.001
Asian American	0.745	0.68 to 0.817	<0.001
Unknown	0.934	0.81 to 1.076	0.34
Obese	1.31	1.214 to 1.415	<0.001

Table 4.

Poisson regression model for incident fracture among boys in the glomerular disease and general pediatric cohorts

Variable	Incidence Rate Ratio	95% CI	P Value
Cohort: glomerular disease	1.148	0.993 to 1.329	0.06
Age (yr)	0.58	0.078 to 4.315	0.59
Age squared	0	0 to 0	<0.001
Age cubed	0.002	0.001 to 0.008	<0.001
Race
White	1	—	—
Black or African American	1.469	1.338 to 1.614	<0.001
Other	1.636	1.507 to 1.777	<0.001
Asian American	0.86	0.759 to 0.973	0.02
Unknown	1.093	0.953 to 1.252	0.2
Obese	1.109	1.036 to 1.188	0.003

Body Site–Specific Incidence

Figure 4 shows the unadjusted incidence rate of fracture at different body sites. As expected, forearm/wrist and hand fractures were the most frequent in both the glomerular disease and general pediatric cohorts. The location-specific fracture SNOMED-CT diagnosis codes with the highest frequency were as follows: “Closed fracture of distal end of radius” (17222009), “Closed fracture of phalanx of finger” (24424003), “Closed supracondylar fracture of humerus” (58580000), “Closed fracture of lower end of radius AND ulna” (33192001), and “Closed torus fracture of radius” (429655000). In analyses of vertebral fractures adjusted for sex, race, age, and obesity, we found that children and young adults with glomerular disease had a five-fold (95% CI, 3.2 to 7.6) increased incidence rate compared with the general pediatric cohort (Supplemental Table 4). Seventeen patients (0.37%) in the glomerular disease cohort had at least one vertebral fracture (unadjusted incidence 8.5 per 10,000 person-years) compared with 335 (0.06%) in the general pediatric cohort (1.0 per 10,000 person-years). For analyses of hip/femur fractures, the glomerular disease cohort had a 2.2-fold (95% CI, 1.3 to 3.7) increased incidence rate (Supplemental Table 5). Ten (0.22%) patients in the glomerular disease cohort had at least one hip/femur fracture (5.8 per 10,000 person-years) compared with 771 (0.14%) in the general pediatric cohort (2.4 per 10,000 person-years).

Figure 4.

Body site–specific fracture incidence in the glomerular disease and general pediatric cohorts.

Incidence of Fracture in the Nephrotic Subcohort Compared with the General Pediatric Cohort

Restricting the glomerular disease cohort to those with nephrotic conditions did not change the overall pattern of results (Supplemental Table 6). The analysis comparing the nephrotic subcohort to the general pediatric cohort suggested that girls in the nephrotic subcohort were 1.6 times more likely (95% CI, 1.3 to 1.9) to experience a fracture, with no such effect for boys.

Incidence of Fracture in Glomerular Disease: Effect of CKD

Among the glomerular disease cohort, 909 (19.8%) patients met the CKD criteria at any time during their follow-up. A model comparing CKD and non-CKD person-time in the glomerular disease cohort to the general pediatric cohort demonstrated that girls with glomerular disease and CKD were 1.9 (95% CI, 1.3 to 2.6) times more likely than the general pediatric cohort to experience a fracture, and for those without CKD, the risk ratio was 1.6 (95% CI, 1.2 to 1.9; Supplemental Table 7). Patients with glomerular disease and CKD were 9.1 (95% CI, 4.9 to 16.9) times more likely than the general pediatric cohort to experience a vertebral fracture, whereas the IRR for those without CKD was 3.5 (95% CI, 2 to 6.3; Supplemental Table 8). For hip/femur fracture, the IRR was 2.2 (95% CI, 0.7 to 6.7) for glomerular CKD and 2.5 (95% CI, 1.5 to 4.3) for glomerular disease without CKD (Supplemental Table 9).

Other Skeletal Outcomes

Figure 5 shows the unadjusted incidence rates of AVN and SCFE for the general pediatric cohort compared with the glomerular disease cohort by CKD and non-CKD person-time.

Figure 5.

AVN and SCFE incidence according to person-time exposed to CKD.

AVN

Thirty (0.65%) patients in the glomerular disease cohort had at least one diagnosis of AVN (unadjusted incidence 26.44 per 10,000 person-years) compared with 66 (0.012%) patients in the general pediatric cohort (0.28 per 10,000 person-years). Adjusted analyses indicated that children with glomerular disease are 56.2 times (95% CI, 40.7 to 77.5) more likely to experience AVN (Supplemental Table 10). The interaction between cohort and sex indicated that the effect of glomerular disease is strengthened in girls and attenuated in boys. The effect of glomerular disease was independent of CKD, with an IRR of 61.9 (95% CI, 36.2 to 105.6) and 50.1 (95% CI, 34.8 to 72.1), respectively, with and without exposure to CKD (Supplemental Table 11).

SCFE

For SCFE, seven (0.15%) patients in the glomerular disease cohort had at least one diagnosis of SCFE (unadjusted incidence 7.6 per 10,000 person-years) compared with 267 (0.048%) patients in the general pediatric cohort (1.4 per 10,000 person-years). Adjusted analyses indicated that children with glomerular disease are 3.4 times more likely (95% CI, 1.9 to 5.9) to experience SCFE (Supplemental Table 12). Obesity was associated with a 10.5-fold (95% CI, 8.4 to 13.2) increased risk. The model that compared glomerular disease CKD and non-CKD person-time to the general pediatric cohort revealed that this effect was driven by CKD. The IRR for glomerular CKD was 10 (95% CI, 5.6 to 18.5), whereas the IRR for glomerular disease without CKD did not reach significance (Supplemental Table 13).

Corticosteroid Exposure Analyses

Contrary to expectations, time-to-event analyses examining time between first glomerular disease diagnosis and first fracture within the glomerular cohort indicated that patients with corticosteroid exposure had a reduced hazard of fracture (hazard ratio=0.6; 95% CI, 0.5 to 0.8; Supplemental Table 14). Within the nephrotic subcohort, steroid exposure did not reach significance for hazard of fracture (Supplemental Table 15). For AVN, as expected, patients with steroid exposure had an increased risk (hazard ratio=2.8; 95% CI, 1.2 to 6.6; Supplemental Table 16).

Sensitivity Analyses

Given the established increased risks of bone loss and fracture in patients with systemic lupus erythematosus (SLE),⁴¹ a sensitivity analysis for fracture that excluded patients with one or more diagnosis codes for systemic lupus erythematosus did not produce a different overall pattern of results (Supplemental Table 17). Of the general pediatric and glomerular disease cohorts, 18.8% and 20.8%, respectively, had more than one fracture. Sensitivity analyses (Supplemental Tables 18–20), which restricted analyses of incidence to each patient’s first fracture event, did not produce a different pattern of results. Similar analyses restricted to the first event of AVN or SCFE also yielded similar results (Supplemental Tables 21 and 22). Finally, a sensitivity analysis for fracture restricting the glomerular disease cohort to the three sites with large regional primary care networks (CHOP, Nemours, and Nationwide) produced similar findings to the main analysis (Supplemental Table 23).

Discussion

Within the population of nearly seven million children in the PEDSnet network, this retrospective cohort study of skeletal outcomes in glomerular disease demonstrated that rates of hip/femur and vertebral fractures, AVN, and SCFE were all at least two-fold higher in the 4598 children with glomerular disease compared with the general pediatric population of 553,624 individuals. There was also a higher risk of all fractures in girls with glomerular disease compared with the reference cohort.

This study leveraged a previously published algorithm that reliably identified children and young adults with glomerular disease and a subcohort of those with nephrotic conditions, with positive predictive values of 89% and 92%, respectively, in our prior study.²⁸ Defining a cohort of children and young adults representative of the general pediatric population from a network of hospital-based health systems poses a challenge due to an overrepresentation of children seeking specialty care. Our approach was to select patients receiving primary care and to exclude patients for whom there was evidence in the EHR of chronic disease. The most frequent visit diagnoses for the general pediatric cohort support the validity of this approach (Supplemental Table 24). Furthermore, the age- and sex-specific patterns of fracture incidence observed in this study among the general pediatric cohort align closely with prior findings for other pediatric studies in the literature across several countries. These include the curvilinear relationship between age and fracture whereby risk is heightened around the time of peak linear growth, an earlier peak in girls than boys, and an increased incidence in boys versus girls particularly accentuated between 10 and 19 years of age.^35,42–47 Consistent with these prior studies, fractures of the forearm/wrist and hand (fingers and metacarpals) were most common, representing 29.8% and 14.6% of all fracture diagnoses, respectively. Obesity was associated with an increased risk of fracture in the reference cohort, also consistent with several previous pediatric studies.^36–38

In comparisons of overall fracture incidence (at any body site) in the glomerular disease and general pediatric cohorts, we found an interaction between sex and cohort. Sex-stratified models indicated that among girls, patients with glomerular disease have an increased risk of fracture compared with the reference cohort—an effect that was not observed in boys. This interaction between sex and glomerular disease was maintained when analyses were restricted to patients who met criteria for the nephrotic subcohort. This sex difference warrants further investigation and could be due to differential effects of glomerular disease on bone quality in girls versus boys. Sex differences in growth-related indices of bone quality, such as greater bone size and strength across growth in boys, may render girls more vulnerable to the effects of glomerular disease.⁴⁸ The sex difference in fracture risk may also be mediated by the effect of glomerular disease on the complex interplay of body composition, pubertal timing, and physical activity on bone accrual.^49–52 Mechanistic studies are needed to determine the deficits in bone structure and strength associated with glomerular disease.

Children with glomerular disease were at increased risk of vertebral and hip/femur fractures. We focused on these fracture sites in particular because they are rare but associated with significant morbidity. In the absence of high energy trauma or falls from a significant height, these fracture sites also indicate underlying bone fragility.^53,54 Furthermore, vertebral fractures may be asymptomatic, and their incidence is likely underestimated. Two prior studies have examined vertebral fractures in children with nephrotic syndrome. One study performed lateral spine radiography in 78 children with nephrotic syndrome within 37 days of corticosteroid initiation and found that 8% had asymptomatic vertebral fractures,⁵⁵ and a prospective cohort study of 54 children with nephrotic syndrome found that 6% experienced vertebral fractures within their first year of corticosteroid therapy.¹⁸

For AVN, which is extremely rare in the general pediatric population,⁵⁶ our analyses indicated that patients with glomerular disease may be at more than a 50-fold increased risk. This is of particular concern because this risk only captures diagnosed AVN. In its early stages, AVN may be associated with only mild symptoms; magnetic resonance imaging is required for adequate detection. For SCFE, our analyses indicate that patients with glomerular disease are at an approximately three-fold greater risk, independent of obesity as a major risk factor for SCFE.

Given the established increased risk of fracture in children with CKD,^13,14 we sought to evaluate how CKD affected the association of glomerular disease with skeletal outcomes. Our analyses suggest that although moderate to advanced CKD increases fracture risk among girls with glomerular disease, the higher fracture risk associated with glomerular disease was still observed in the absence of CKD. Similarly, the higher risks of vertebral and hip/femur fractures and AVN were also independent of CKD. In contrast, the increased SCFE risk among the glomerular disease cohort was driven by those with CKD.

In time-to-event analyses within the cohort of children with glomerular disease and in the subcohort with nephrotic conditions, we did not find the expected association of corticosteroid exposure with fracture risk.^57,58 This unexpected result is likely due to confounding effects by other associated patient characteristics (e.g., an inverse relationship between corticosteroid exposure and physical activity or pattern of steroid responsiveness). Albeit unexpected, this finding highlights that the skeletal risk in this patient population is not attributable to corticosteroid exposure alone and underscores the importance of addressing other modifiable risk factors. Time-to-event analyses demonstrating that corticosteroid exposure was associated with increased hazard of AVN were consistent with previous literature.⁵⁶

This study has several limitations, some of which are inherent in the use of multi-institutional EHR data resources. The data are limited to care received within the PEDSnet network. If, for example, a patient received care for a fracture at an urgent care facility outside the network, this fracture would not necessarily be captured by this approach. This limitation poses a particular concern if patients in one of our cohorts of interest were more likely to seek care within the network for reasons of geography or increased clinical concern. We attempted to mitigate this issue by requiring evidence of consistent follow-up in our cohort definitions. If a single patient received care at multiple institutions, they would be represented as a separate individual at each site. For children with a chronic health condition, such assignment may occur due to “second opinion” visits. We attempted to mitigate against this issue by requiring 12 months of clinical follow-up. This frequency of required follow-up may have selected for more “sick” visits and less representativeness of the healthy pediatric population but would likely have biased our results toward the null. Because our intention was to identify a cohort broadly representative of the general pediatric population in the network, our application of the PMCA algorithm did not exclude all chronic diseases from this cohort. Although infrequent, the inclusion of patients with chronic conditions, such as inflammatory bowel disease or diabetes, would have biased our results toward the null. Although the classification accuracy of the computable phenotype used to identify glomerular disease was previously evaluated in PEDSnet with a positive predictive value >85%, we assume there was misclassification of exposure likely by other forms of kidney disease. Given the inherent limitations of secondary use of EHR data, we were unable to categorize the patients with nephrotic conditions according to their steroid responsiveness.

Our approach to outcome ascertainment could be sensitive to timing of the outcome relative to a patient’s birthday, creating some misclassification with respect to age or double counting of an event proximate to a birthday, although we expect this would be nondifferential between the cohorts. Exposure to corticosteroids was based on number of prescriptions and not cumulative dose or duration. Some covariates of interest were either unavailable or incompletely captured for analysis, such as 25-hydroxyvitamin D levels, which are not routinely measured in the healthy population, and pubertal stage, measures of physical activity, and sports participation, which are not well recorded in the EHR. Although the cohort size of more than 4500 children and young adults with glomerular disease is large, given the rarity of these conditions, we still had limited power to investigate multiplicative interactions for rare outcome events such as hip and femur fractures and SCFE. Our analytical approach to fracture, AVN, and SCFE incidence did not account for potential correlation at the patient level if a patient had more than one event during their follow-up period. Sensitivity analyses examining incidence of first events for fracture, AVN, and SCFE were, however, consistent with the primary analyses.

In summary, these data show that children with glomerular disease are at increased risk for fractures of the vertebra and hip/femur, AVN, and SCFE compared with a general pediatric reference cohort. For fractures at any body site, girls with glomerular disease are at increased risk compared with girls in the general pediatric population. Our findings also add to a literature that has demonstrated an association between nephrotic syndrome and reduced bone mineral density.^15–27 This large retrospective cohort study of skeletal outcomes in children and young adults with glomerular disease highlights the strength of multi-institutional EHR data for studying outcomes in pediatric rare diseases. Studies are needed to address mechanisms and prevention of skeletal complications in this population.

Disclosures

L.C. Bailey reports research funding from Jazz Pharmaceuticals, Takeda Pharmaceuticals, and UCB Pharmaceuticals. C. Gluck reports honoraria from Retrophin and Sanofi Genzyme, and an advisory or leadership role for Retrophin and Sanofi Genzyme. B.P. Dixon reports consultancy for Alexion Pharmaceuticals and Apellis Pharmaceuticals, and honoraria from Alexion Pharmaceuticals and Apellis Pharmaceuticals. J.T. Flynn reports honoraria from Springer (royalties) and UpToDate (royalties); an advisory or leadership role for Blood Pressure Monitoring (editorial board member), Hypertension (editorial board member), Journal of Pediatrics (editorial board member), Pediatric Nephrology (editor-in-chief), and the Renal Physicians Association (board member); and other interests or relationships with the American Society of Pediatric Nephrology, International Pediatric Nephrology Association, and Renal Physicians Association. D. Ranade reports an advisory or leadership role for PEDSnet (research subcommittee chair). W.E. Smoyer reports consultancy for Otsuka, Vertex, and Visterra; ownership interest in NephKey Therapeutics; research funding from Aurinia; honoraria from UCLA–CTSA External Advisory Committee and USC–CTSA External Advisory Committee; patents or royalties from UpToDate; and an advisory or leadership role for the Institute for the Advancement of Clinical Trials in Children (member of the coordinating committee), Kidney International, NephCure (member of the board of directors), and the Pediatric Nephrology Research Consortium (member of the board of directors). V.R. Dharnidharka reports consultancy for Atara Biotherapeutics and Medincell; research funding from CareDx; honoraria from CareDx; an advisory or leadership role for the North American Pediatric Renal Trials and Collaborative Studies; and other interests or relationships with Akebia/MedPace and the Independent Data Safety Monitoring Committee. B. Magnusen reports being an employee of AT&T, and consultancy for Bayer, Daugherty Business Solutions, and The Climate Corporation. M.M. Mitsnefes reports consultancy for KBP Biosciences. M. Somers reports consultancy for Alnylam, Dicerna, Horizon, and Orfan Biotech, and an advisory or leadership role for ASPN (president) and NAPRTCS (board of directors). E.K. Burrows reports being an employee of Helix, and ownership interest in Helix. C.B. Forrest reports research funding to institution from Bayer, Lily, Sanofi, and UCB, and patents or royalties from Johns Hopkins University. M.R. Denburg reports research funding from Mallinckrodt; an advisory or leadership role for Kidney International Reports (editorial board member) and the NKF Delaware Valley Medical Advisory Board; and other interests or relationships with the American Society of Pediatric Nephrology Research and Program Committees and the National Kidney Foundation Pediatric Education Planning Committee. All remaining authors have nothing to disclose.

Funding

This study was funded by National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) grant R21DK116151 (principal investigator: M.R. Denburg). This work was also supported in part by The Patient-Centered Outcomes Research Institute grant award RI-CHOP-01-PS1 and NIDDK grant P50DK114786.

Author Contributions

L.C. Bailey, E.K. Burrows, S.J. Deakyne Davies, D. Eckrich, C.B. Forrest, A.J. Goodwin Davies, M. Kitzmiller, B. Magnusen, D. Ranade, H. Razzaghi, and L. Utidjian were responsible for data curation; L.C. Bailey, D.J. Claes, M.R. Denburg, V.R. Dharnidharka, B.P. Dixon, J.T. Flynn, C.B. Forrest, S.L. Furth, C. Gluck, M.M. Mitsnefes, W.E. Smoyer, and M. Somers were responsible for the investigation; L.C. Bailey, M.R. Denburg, C. Forrest, A.J. Goodwin Davies, H. Razzaghi, and R. Xiao were responsible for the formal analysis; L.C. Bailey, M.R. Denburg, C.B. Forrest, A.J. Goodwin Davies, H. Razzaghi, R. Xiao, and L. Utidjian were responsible for the methodology; L.C. Bailey, M.R. Denburg, H. Razzaghi, and R. Xiao were responsible for supervision; M.R. Denburg and C.B. Forrest were responsible for conceptualization and funding acquisition; M.R. Denburg, C.B. Forrest, and I.Y. Luna were responsible for project administration; M.R. Denburg, A.J. Goodwin Davies, H. Razzaghi, and R. Xiao wrote the original draft of the manuscript; A.J. Goodwin Davies was responsible for visualization; and all authors reviewed and edited the manuscript.

Supplemental Material

This article contains the following supplemental material online at http://jasn.asnjournals.org/lookup/suppl/doi:10.1681/ASN.2021101372/-/DCSupplemental.

Supplemental Table 1. Summary of all analyses performed.

Supplemental Table 2. Poisson regression model for incident fracture in the general pediatric cohort.

Supplemental Table 3. Overall Poisson regression model for incident fracture in the glomerular disease and general pediatric cohorts (not stratified by sex).

Supplemental Table 4. Poisson regression model for incident vertebral fractures.

Supplemental Table 5. Poisson regression model for incident hip and femur fractures.

Supplemental Table 6. Poisson regression model for incident fracture in the nephrotic subcohort and general pediatric cohort.

Supplemental Table 7. Poisson regression model for incident fracture in the glomerular and general pediatric cohorts, evaluating person-time exposed to chronic kidney disease (CKD) in the glomerular cohort.

Supplemental Table 8. Poisson regression model for vertebral fracture, evaluating person-time exposed to chronic kidney disease (CKD) in the glomerular cohort.

Supplemental Table 9. Poisson regression model for hip and femur fractures, evaluating person-time exposed to chronic kidney disease (CKD) in the glomerular cohort.

Supplemental Table 10. Poisson regression model for avascular necrosis/osteonecrosis (AVN).

Supplemental Table 11. Poisson regression model for avascular necrosis/osteonecrosis (AVN), evaluating person-time exposed to chronic kidney disease (CKD) in the glomerular cohort.

Supplemental Table 12. Poisson regression model for slipped capital femoral epiphysis (SCFE).

Supplemental Table 13. Poisson regression model for slipped capital femoral epiphysis (SCFE), evaluating person-time exposed to chronic kidney disease (CKD) in the glomerular cohort.

Supplemental Table 14. Cox proportional hazards model evaluating the association between corticosteroid exposure and time to first fracture within the glomerular disease cohort.

Supplemental Table 15. Cox proportional hazards model evaluating the association between corticosteroid exposure and time to first fracture within the nephrotic subcohort.

Supplemental Table 16. Cox proportional hazards model evaluating the association between corticosteroid exposure and time to avascular necrosis/osteonecrosis (AVN) within the glomerular disease cohort.

Supplemental Table 17. Poisson regression model for incident fracture in the glomerular disease (excluding the 448 patients with one or more lupus diagnosis codes) and general pediatric cohorts.

Supplemental Table 18. Poisson regression model for first fracture in the glomerular disease and general pediatric cohorts.

Supplemental Table 19. Poisson regression model for first vertebral fracture in the glomerular disease and general pediatric cohorts.

Supplemental Table 20. Poisson regression model for first hip/femur fracture in the glomerular disease and general pediatric cohorts.

Supplemental Table 21. Poisson regression model for first diagnosis of avascular necrosis/osteonecrosis (AVN) in the glomerular disease and general pediatric cohorts.

Supplemental Table 22. Poisson regression model for first diagnosis of slipped capital femoral epiphysis (SCFE) in the glomerular disease and general pediatric cohorts.

Supplemental Table 23. Overall Poisson regression model for incident fracture in the glomerular disease and general pediatric cohorts (not stratified by sex), limiting the glomerular disease cohort to patients at sites with a primary care network.

Supplemental Table 24. The most frequent visit diagnoses for the general pediatric cohort (ranked by number of patients).