Journal Logo

Scientific Articles

PPARGC1B Is Associated with Nontraumatic Osteonecrosis of the Femoral Head

A Genomewide Association Study on a Chart-Reviewed Cohort

Zhang, Yanfei MD, PhD1,2; Bowen, Thomas R. MD2; Lietman, Steven A. MD2; Suk, Michael MD2; Williams, Marc S. MD1; Lee, Ming Ta Michael PhD1,2,a

Author Information
The Journal of Bone and Joint Surgery: September 16, 2020 - Volume 102 - Issue 18 - p 1628-1636
doi: 10.2106/JBJS.19.01335
  • Open
  • Supplementary Content
  • Disclosures


Osteonecrosis of the femoral head (ONFH) is a disabling disease primarily affecting adults in their 30s to 50s1,2. This condition develops when the circulation to the femoral head is critically impaired. Major trauma that disrupts the vascular system can lead to osteonecrosis. However, the exact mechanism of vascular disruption is unknown for nontraumatic ONFH. ONFH is relatively common in the U.S., affecting approximate 20,000 to 30,000 people per year3. If untreated, ONFH progressively impairs the ability to walk. Numerous treatments have been attempted, but efforts at hip preservation often fail. Currently, there is no accepted treatment guideline for ONFH4. Moreover, patients routinely present late in the disease process when secondary osteoarthritis of the hip is severe. Consequently, hip arthroplasty is the only effective option for these patients despite their young age and active lifestyles3,5.

Risk factors for ONFH include alcohol consumption, (cortico)steroid treatment, sickle-cell disease, Legg-Calvé-Perthes disease, Gaucher disease, autoimmune diseases, coagulation disorders, hemoglobinopathies, HIV/AIDS (human immunodeficiency virus/acquired immune deficiency syndrome), pancreatitis, organ transplantation, myeloma, and leukemia4. The predominant risk factors in the U.S. are alcohol consumption and steroid treatment4. Previous studies suggested that genetic variation is a contributing factor6-12. ONFH is frequently bilateral, which also implicates a genetic influence. Candidate-gene studies have reported variants in genes affecting the coagulation pathway, angiogenesis, inflammation, and drug metabolism11. Recent genomewide association studies (GWASs) (see Appendix) identified genetic variants associated with all types of ONFH: steroid and alcohol-associated ONFH in adults9,10, steroid-associated ONFH in children with acute lymphoblastic leukemia (ALL)7,8, and nontraumatic ONFH in adults10,12. Most studies support a “multiple-hit” model of ONFH pathogenesis, in which numerous subclinical factors, including genetic predisposition, combine to reach a critical threshold of bone ischemia followed by clinically apparent ONFH3,4,13.

In this study, we tested the hypothesis that other unidentified genetic variants contribute to the pathophysiology of ONFH and that elucidation of these factors could reveal etiologic pathways. We performed a GWAS to screen the genomes of patients with clinically and radiographically documented nontraumatic ONFH in comparison with a local reference population exposed to similar environmental influences but without ONFH.

Materials and Methods

The study cohort consisted of participants from the MyCode Community Health Initiative, a large population-based research program in which Geisinger patients provide written consent to allow their clinical and genomic data to be used for research14,15. This study received approval by the Geisinger institutional review board (IRB #2017-0345) for chart and image review, with a waiver to use deidentified data, and performance of a genetic study was approved by the MyCode governing board.

Chart Review and Case and Control Definitions

Figure 1 illustrates the study workflow. Patients who had an ICD (International Classification of Diseases) code for ONFH (ICD-9: 733.42; ICD-10: M87.050, M87.051, M87.052, or M87.059) were identified from electronic health records (EHRs). An experienced orthopaedic surgeon (>10 years in the ONFH field) performed the chart and radiographic review to confirm the nontraumatic ONFH diagnosis. In the cohort, 75.7% of patients had computed tomography (CT) scans or magnetic resonance imaging (MRI) and 96.4% had radiographs; thus, most patients had both a radiograph and either a CT scan or MRI. As radiography is less sensitive for detecting early abnormality, we investigated the 34 patients with only a radiograph. Thirty-two had collapsed ONFH, another had ONFH documented in a CT report although a CT image was not available in the record, and the last patient had 3 encounters during 4 years in which bilateral ONFH was documented in the charts. Patients with osteonecrosis in other joints or traumatic osteonecrosis were excluded. Population controls were identified by excluding patients with osteonecrosis of any joint on the basis of ICD codes.

Fig. 1
Fig. 1:
Study flowchart. ONFH cases and controls were identified from MyCode participants based on ICD codes. Chart review was then performed to identify true nontraumatic ONFH cases based on radiographs and MRI. The risk of ONFH and confounding variables were extracted from the deidentified EHR database based on the ICD and CPT (Current Procedural Terminology) codes. Genetic analysis only included individuals who had European ancestry and were at third-degree or more distant relatedness. GWAS was performed on all cases versus all controls, followed by a gene-based analysis using MAGMA and sensitivity analyses on a steroid subset and on ONFH cases with a collapsed femoral head. T2DM = type-2 diabetes mellitus, SLE = systemic lupus erythematosus, HIV = human immunodeficiency virus, and BMI = body mass index. M87.05* includes subcodes M87.050, M87.051, M87.052, and M87.059.

Data Deidentification and EHR Data Extraction

The chart-reviewed data set was deidentified and linked to the genetic data by the Geisinger Phenomic Analytics and Clinical Data Core. Age, sex, risk factors, diseases (see Appendix Table 1), and steroid use were extracted from the EHR. Steroid exposure was defined as >3 weeks of prednisone and/or dexamethasone intake by mouth, injection, or infusion; steroid inhalers were excluded.

Genetic Analyses and in Silico Prediction

Genotyping and quality control of the imputed genetic data were performed as previously described (see Appendix)16. Common variants (minor allele frequency [MAF], >1%) and unrelated individuals were included in the analyses. A total of 118 patients with ONFH, 56,811 controls, and 3,905,175 single nucleotide variants (SNVs) passed the quality control criteria. A logistic regression model was adjusted for significant covariates including age, sex, clinical risk factors, and genetic principal components under an additive genetic mode using PLINK version 1.9 ( MAGMA (version 1.07b; was used to perform gene-based association tests18, using the parameters “snp-wise” and “top 15%.” SNVs were mapped to 17,614 genes. A p value of <2.84 × 10−6 (0.05/17,614) was considered significant under Bonferroni correction. Because steroid exposure is the most common risk factor for ONFH, and collapsed ONFH represents the severe phenotype, we performed sensitivity analyses for (1) the top variants in patients with steroid-associated ONFH versus controls with steroid intake (steroid subset), and (2) patients with ONFH and a collapsed femoral head versus all controls (collapsed subset), using logistic regression adjusting for significant covariates. In silico annotation and prediction included the following software: STRING ( to identify protein-protein interaction networks; GTEx portal (; V8 release) to query the eQTL (expression quantitative trait loci), sQTL (splicing quantitative trait loci), and expression profiles in different tissue types; and Ensemble ( to annotate (Variant Effect Predictor [VEP], and visualize the variants, gene structure, and motifs.

Statistical Analyses

The chi-square test was used for comparison of categorical variables, and 1-way analysis of variance (ANOVA) was used to compare continuous variables. P < 0.05 was considered significant. Analyses were performed using R (version 3.5.1; R Foundation for Statistical Computing). To adjust the multiple tests, the genomewide significance level was defined as p < 5 × 10−8, and a suggestive significance level was defined at p < 1 × 10−5. A power analysis for the GWAS was conducted using Quanto (, assuming a suggestive significance level of p < 1 × 10−5 and an ONFH prevalence of 0.01%. The current sample size would yield 80% power for variants with an MAF of 2% and an odds ratio (OR) of ≥4.5, or with an MAF of 5% and OR of ≥3.1.


Chart Review and Cohort Characterization

Of the 190 patients who had ICD codes for ONFH, 6 patients did not have imaging studies available for review, 33 patients had ONFH incorrectly coded, and 11 patients had ONFH associated with trauma (Fig. 1). The remaining 140 patients with radiographically confirmed nontraumatic ONFH formed the study cohort. The error rate for ICD code-based identification of nontraumatic ONFH was 26.3%. No patients with ONFH had Gaucher disease or Legg-Calvé-Perthes disease; therefore, we excluded controls with these 2 diseases. We included 118 patients with nontraumatic ONFH and 56,811 controls after quality control of the genetic data. The demographics and clinical characteristics of the cohort are summarized in Table I. Most of the patients with ONFH were male (69.5%). Alcohol abuse, HIV infection, systemic lupus erythematosus, leukemia, and steroid intake were significantly associated with ONFH in the patient cohort. Steroid intake was highly prevalent in both cases (90.7%) and controls (68.3%). Other risk factors such as pancreatitis, organ transplantation, and lymphoma were more prevalent in the patients with ONFH than in the controls, although the differences did not reach significance.

TABLE I - Characteristics of the ONFH Cases and Controls Used for the Genetic Study*
ONFH (N = 118) Controls (N = 56,811) P Value
Female sex 36 (30.5%) 33,609 (59.2%) 4.7 × 10−10
Age(yr) 50.6 ± 13.3 59.2 ± 17.4 1.6 × 10−10
Alcohol abuse 28 (23.7%) 2,555 (4.5%) <2.2 × 10−16
HIV 3 (2.54%) 92 (0.16%) 2.0 × 10−7
Pancreatitis 3 (2.54%) 963 (1.7%) 0.72
Organ transplant 2 (1.69%) 439 (0.78%) 0.54
Systemic lupus erythematosus 6 (5.08%) 372 (0.65%) 8.7 × 10−8
Leukemia 3 (2.54%) 354 (0.62%) 0.04
Lymphoma 3 (2.54%) 517 (0.91%) 0.17
Type-2 diabetes 22 (18.6%) 14,022 (24.7%) 0.16
Steroid 107 (90.7%) 38,791 (68.3%) 3.0 × 10−7
*ONFH = osteonecrosis of the femoral head.
The values are given as the mean and standard deviation.
Includes only prednisone or dexamethasone, given orally or by injection or infusion for >3 weeks.

GWAS-Identified Significant Loci Associated with Nontraumatic ONFH

Figures 2-A and 2-B show a Manhattan plot and QQ (quantile-quantile) plot for the GWAS. Table II lists the 4 independent loci that reached genomewide significance (p < 5 × 10−8) and another 20 independent loci with suggestive significance (p < 1 × 10−5). The most significant variant was rs116468452, an intergenic SNV on chromosome 1q25.3. Of the remaining 3 loci with genomewide significance, rs138371993 and rs111529765 are also in the intergenic regions, and rs112517560 is an intron variant of SYN2 (synapsin-2). We did not find any significant interactions between these 4 top SNVs and confounding variables (p > 0.05).

Fig. 2
Fig. 2:
Figs. 2-A through 2-D GWAS and gene-based analysis for all cases and controls. Figs. 2-A and 2-B Manhattan plot and QQ plot of the GWAS for all ONFH cases versus all controls. The blue and red lines represent a suggestive significance level (p = 1 × 10−5) and the genomewide significance level (p = 5 × 10−8), respectively. Fig. 2-C Manhattan plot of gene-based analysis using summary statistics of the GWAS for all ONFH cases versus all controls. The red line represents the Bonferroni-corrected significance level (p = 2.84 × 10−6). Fig. 2-D Protein-protein interaction network for PPARGC1B identified by STRING. The nodes and edges, respectively, represent encoded proteins and evidence-based functional interactions derived from a combination of resources including experiments, curated databases, coexpression, and text-mining of publications. Only high-confidence interactions (score, >0.9) are shown.
TABLE II - Independent Variants with P < 10−5 in the GWAS for All Patients with ONFH, and Their Associations in the Steroid and Collapsed Subsets*
SNV CHR:BP A1/A2 Gene MAF All Cases and Controls Steroid Subset Collapsed Subset
OR (95% CI) P Value OR (95% CI) P Value OR (95% CI) P Value
rs116468452 1:181930202 C/T 0.011 5.51 (3.13, 9.69) 3.26 × 10−9 5.60 (3.21, 9.76) 1.30 × 10−9 6.12 (3.32, 11.27) 6.53 × 10−9
rs138371993 4:52971435 G/A 0.012 5.47 (3.02, 9.89) 1.93 × 10−8 4.75 (2.49, 9.05) 2.25 × 10−6 5.83 (3.05, 11.14) 9.19 × 10−8
rs111529765 7:15898188 T/C 0.031 3.40 (2.21, 5.23) 2.51 × 10−8 3.15 (2.00, 4.96) 7.13 × 10−7 3.44 (2.11, 5.60) 6.90 × 10−7
rs112517560 3:12106685 G/A SYN2 0.012 5.32 (2.95, 9.60) 2.92 × 10−8 4.48 (2.36, 8.48) 4.32 × 10−6 4.59 (2.24, 9.39) 3.02 × 10−5
rs2189071 7:88784230 T/A ZNF804B 0.013 5.01 (2.76, 9.10) 1.26 × 10−7 4.16 (2.18, 7.96) 1.64 × 10−5 4.60 (2.31, 9.14) 1.34 × 10−5
rs10953090 7:92728528 A/G SAMD9 0.092 2.38 (1.72, 3.30) 1.87 × 10−7 2.57 (1.84, 3.58) 2.96 × 10−8 2.56 (1.79, 3.67) 3.19 × 10−7
rs7825223 8:104113149 A/G 0.011 4.95 (2.66, 9.23) 4.54 × 10−7 4.46 (2.35, 8.47) 4.72 × 10−6 5.11 (2.57, 10.17) 3.29 × 10−6
rs3808749 9:17622808 A/G SH3GL2 0.028 3.25 (2.03, 5.20) 8.89 × 10−7 3.38 (2.09, 5.47) 6.68 × 10−7 2.84 (1.62, 4.97) 2.58 × 10−4
rs77249562 22:27759273 G/A CTA-929C8.8 0.041 2.90 (1.89, 4.47) 1.33 × 10−6 2.35 (1.46, 3.77) 4.22 × 10−4 3.10 (1.93, 4.99) 2.99 × 10−6
rs78814834 5:149191547 T/C PPARGC1B 0.042 2.86 (1.87, 4.38) 1.40 × 10−6 2.87 (1.85, 4.45) 2.40 × 10−6 3.09 (1.94, 4.92) 2.24 × 10−6
rs145185972 8:365972 G/C FBXO25 0.023 3.43 (2.07, 5.69) 1.79 × 10−6 3.78 (2.29, 6.25) 2.10 × 10−7 3.70 (2.11, 6.46) 4.48 × 10−6
rs57220573 16:7408636 C/G RBFOX1 0.011 4.36 (2.37, 8.01) 2.17 × 10−6 4.44 (2.35, 8.39) 4.55 × 10−6 3.03 (1.35, 6.80) 7.42 × 10−3
rs761440 20:24648764 A/G RP5-860P4.2 0.016 3.92 (2.23, 6.90) 2.24 × 10−6 3.27 (1.79, 6.00) 1.25 × 10−4 4.16 (2.25, 7.70) 5.52 × 10−6
rs584709 6:154419539 C/T OPRM1 0.341 1.86 (1.44, 2.40) 2.35 × 10−6 1.75 (1.33, 2.28) 4.92 × 10−5 1.90 (1.43, 2.54) 1.35 × 10−5
rs73226405 4:10477165 T/G RP11-136I13.2 0.094 2.25 (1.60, 3.17) 3.40 × 10−6 2.19 (1.53, 3.12) 1.60 × 10−5 2.12 (1.43, 3.15) 1.84 × 10−4
rs11721852 4:7346719 T/C SORCS2 0.030 3.04 (1.90, 4.87) 3.92 × 10−6 3.00 (1.85, 4.86) 8.00 × 10−6 2.59 (1.47, 4.55) 9.66 × 10−4
rs114703252 5:56485428 A/T GPBP1 0.016 3.64 (2.09, 6.35) 5.12 × 10−6 3.48 (1.93, 6.26) 3.25 × 10−5 4.46 (2.50, 7.96) 4.25 × 10−7
rs10236649 7:84992825 C/T 0.359 1.82 (1.41, 2.36) 5.15 × 10−6 1.77 (1.36, 2.32) 2.85 × 10−5 1.99 (1.49, 2.66) 3.79 × 10−6
rs149415799 1:173657022 C/G 0.019 3.41 (2.01, 5.77) 5.44 × 10−6 3.79 (2.26, 6.37) 4.85 × 10−7 2.63 (1.34, 5.14) 4.85 × 10−3
rs61811251 1:180891069 A/G KIAA1614 0.011 4.46 (2.34, 8.52) 5.93 × 10−6 4.07 (2.08, 8.00) 4.42 × 10−5 3.97 (1.85, 8.52) 4.13 × 10−4
rs55863623 3:34879269 A/G 0.027 3.08 (1.90, 5.01) 6.29 × 10−6 3.35 (2.07, 5.43) 9.55 × 10−7 3.58 (2.13, 6.01) 1.49 × 10−6
rs117234581 17:74323603 A/G PRPSAP1 0.018 3.46 (2.01, 5.95) 7.75 × 10−6 3.79 (2.21, 6.48) 1.23 × 10−6 3.52 (1.91, 6.49) 5.36 × 10−5
rs112467115 4:25172559 T/C PI4K2B 0.020 3.42 (1.99, 5.87) 8.42 × 10−6 3.19 (1.80, 5.67) 7.15 × 10−5 4.50 (2.60, 7.79) 7.82 × 10−8
rs7771691 6:154820852 C/T CNKSR3 0.121 2.05 (1.50, 2.81) 8.50 × 10−6 2.01 (1.45, 2.79) 2.96 × 10−5 1.87 (1.29, 2.70) 8.65 × 10−4
*ONFH = osteonecrosis of the femoral head, SNV = single nucleotide variant, CHR:BP = chromosome location and coordinates in base pairs, A1 = risk allele (which is also the minor allele), A2 = reference allele, MAF = minor allele frequency, OR = odds ratio, and CI = confidence interval.

We also tested the previously reported ONFH-associated SNV rs10989692 in GRIN3A (OR = 0.4, p = 0.0079); however, it showed an opposite effect from that described in the prior study7. We did not observe significant associations for other variants from prior studies (see Appendix Tables 2 and 3).

Sensitivity Analyses of the Top Variants in the Steroid and Collapsed ONFH Subsets

The effects of the top variants remained in the same direction in the sensitivity analyses involving the steroid subset and the collapsed subset (Table II). rs116468452 remained genomewide significant in both subsets (p = 1.3 × 10−9 and 6.53 × 10−9, respectively). rs10953090 in SAMD9 became significant in the steroid subset (p = 2.96 × 10−8). rs112467115 in PI4K1B showed a stronger association with a larger effect size in the collapsed subset (p = 7.82 × 10−8, OR = 4.5) than in the entire cohort.

PPARGC1B Is Associated with Nontraumatic ONFH

Gene-based analysis identified PPARGC1B (peroxisome proliferator-activated receptor gamma, coactivator 1 beta) as the only significant gene after Bonferroni correction (p = 1 × 10−6, Fig. 2-C). A protein-protein interaction network for PPARGC1B is shown in Figure 2-D. PPARGC1B can bind to several transcription factors and nuclear receptors, including the glucocorticoid receptor NR3C1 (nuclear receptor subfamily 3 group C member 1); the steroid hormone receptors ESRRA (estrogen-related receptor alpha), ESR1 (estrogen receptor 1), and ESRRG (estrogen-related receptor gamma); NRF1 (nuclear respiratory factor 1); and PPARGC1A (PPARG coactivator 1 alpha). This network is highly enriched in nuclear receptor activity (false discovery rate [FDR] = 7.3 × 10−9) and steroid hormone receptor activity (FDR = 1.2 × 10−6).

A plot of the PPARGC1B region with annotation of the variants is shown in Figures 3-A and 3-B. SNVs in PPARGC1B with a suggestive significance (p < 1 × 10−5) are listed in Appendix Table 4. GTEx revealed that none of these variants are eQTLs or sQTLs in any currently studied tissue types. Most variants are intronic, including the lead SNV rs78814834. The missense variant rs45520937 is in high linkage disequilibrium (R2 = 0.945, p = 3.7 × 10−6) with the lead SNV, causing R/Q substitution, which is predicted to be benign, in several transcripts of PPARC1B. Seven variants are located at the CTCF (CCCTC-binding factor) binding sites. The expression of 7 isoforms of PPARGC1B is shown in Appendix Figure 1. The ENST00000309241.9 isoform is more widely expressed in most tissue types than the other 6. However, to our knowledge, information on bone-specific isoforms is not available.

Fig. 3
Fig. 3:
Figs. 3-A and 3-B The association of the PPARGC1B locus with ONFH. SNP = single nucleotide polymorphism, a synonym for SNV. Fig. 3-A Regional association of the PPARGC1B locus with ONFH. The lead SNV rs78814834 (dark purple dot) is labeled. Fig. 3-B The annotation of the SNVs in the PPARGC1B locus. The chromosome bands, SNVs (VEP result), gene transcripts (GENCODE19), and regulatory build are shown. The upper panel shows an overview of the PPARGC1B locus, the middle panel shows a zoomed-in view of the SNVs, and the legend in the lower panel shows variant and regulation information.


Our study using EHR and genomic data from a large population cohort identified several SNVs and 1 gene associated with ONFH. The EHR encompasses multidimensional data, providing ample opportunities for genetic studies but also challenges, such as data fidelity, especially for many skeletal diseases. Diagnosis of skeletal diseases relies heavily on radiographic interpretations, which are poorly presented in the EHR. By performing radiographic image review, we observed a high coding error rate of 26.3%. Many scenarios can lead to incorrect coding of ONFH. For example, patients with other skeletal disorders (such as osteoarthritis of the hip or osteonecrosis of other joints) may have had an ONFH ICD code entered into the EHR system to justify radiographic examinations, but subsequent radiographs revealed no ONFH. The strength of our study is the fidelity of the phenotype based on careful image and chart review, which is essential for excluding miscoded cases and traumatic ONFH to yield a precisely defined group.

Pathways involving steroid metabolism, metabolic disorders, circulation, chronic inflammation, and bone regulation are commonly acknowledged in the etiology and pathogenesis of ONFH11. Risk factors associated with these pathways, such as alcohol consumption, steroid intake, HIV/AIDS, systemic lupus erythematosus, and leukemia, were found at higher prevalence in patients with ONFH. Our study identified SNVs that may reveal novel ONFH mechanisms. The intergenic SNV rs116468452 was consistently and significantly associated with ONFH. The nearest gene, CACNA1E, encodes calcium voltage-gated channel subunit alpha-1E, a member of the high voltage-activated R-type calcium channel family. CACNA1E was overexpressed fivefold in the tissue from ONFH bone samples compared with normal bone, providing independent support for the significance of the association19. rs10953090, the significant SNV in the steroid subset, is located downstream of SAMD9. Mutations in SAMD9 can cause normophosphatemic familial tumoral calcinosis20,21 and MIRAGE (myelodysplasia, infection, restriction of growth, adrenal hypoplasia, genital phenotypes, and enteropathy) syndrome, which can lead to myelodysplastic syndrome or acute myeloid leukemia (AML)22-25. While these diseases are associated with increased risk of ONFH, the mechanism for ONFH in the absence of myelodysplasia or AML remains to be elucidated. rs112467115 in PI4K2B and rs114703252 in GPBP1 displayed stronger associations with collapsed ONFH. Both genes may be involved in thrombosis, a known factor of ONFH etiology. PI4K2B, together with PI4K2A, regulate the formation of Weibel-Palade bodies, which mediate the release of molecules involved in thrombosis, inflammation, and angiogenesis26. GPBP1, encoding for vasculin, has been found to be expressed in atherosclerotic plaques with a thrombus27.

Gene-based analyses provide better statistical power by aggregating SNV data at the gene level. PPARGC1B was the only significant gene in the gene-based analyses. A recent GWAS of nontraumatic ONFH identified a significant gene, PPARGC, which also mediated the increased risk of patients treated with thiazolidinediones (TZDs), which are antidiabetic drugs12. Protein-protein interaction analysis predicted an interaction between PPARGC1B and PPARGC with high confidence, further strengthening the evidence for the association of this pathway with ONFH. Genetic studies have also identified SNVs in PPARGC1B associated with type-2 diabetes, adipogenesis, insulin resistance, obesity, and fasting glucose levels28-31. These findings, considered together with our study, provide further evidence regarding the metabolic etiology of ONFH and suggest that genes in this common pathway play important roles in ONFH.

PPARGC1B can regulate the osteoclast cytoskeleton by mitochondrial biogenesis and activation32 and stimulate the transcription of the estrogen receptor33,34. In our study, we observed a large sex difference in ONFH (2.3:1 male:female ratio), consistent with previous observations35-37. Although this may be partially due to the tendency for higher alcohol consumption in men, the estrogen hormone and its receptor may also play a critical role, especially given their important role in thrombosis, which can induce ischemia of the femoral head. Studies have found that males have a twofold higher risk than females of developing venous thromboembolism38,39. Variants in PPARGC1B are associated with thromboxane A2 formation40, providing further evidence for its role in thrombosis. The intronic variants of the PPARGC1B locus, including the lead SNV rs78814834, and variants located in the 3′ UTR (untranslated) region and CTCF binding sites, may regulate splicing, RNA stability, and gene expression.

The study has several limitations. First, like previous studies, the sample size of ONFH cases is small (118 patients), and the study is underpowered for variants with a low frequency or a small effect size (OR, <3). Because chart review was not performed in the control group, there may also have been cases of ONFH in that group. However, given the low prevalence of ONFH, the broad exclusion codes used for selection of the control group, and the large sample size of the controls, this potential weakness may not substantially affect the results. Second, no other cohorts with radiographically confirmed ONFH were available to validate our results. The results of prior genetic association studies of ONFH have also been poorly replicated, and we were not able to validate the previously reported ONFH-associated variants. The reasons could be small sample sizes, heterogeneity in the studied phenotypes and ethnic groups, and different case definitions, with the undetermined misclassification rate being a weakness in some of the prior studies. On the other hand, the lack of replication supports the complexity of the polygenic background and multifactorial etiology of ONFH, which involves multiple genes, variable phenotypic penetrance of the variants, and complex gene-environment interactions. Third, we were not able to perform functional validation of the variants in PPARGC1B to elucidate their mechanistic effects, such as bone-specific splicing or changes in expression level. Future studies may focus on the functional validation of the identified variants using cell, tissue culture, and mouse models. Tissue banking by clinical surgeons can provide valuable resources to examine the clinical functional impact of these variants. Moreover, the interaction of PPARGC1B genotype with TZD treatment in patients with diabetes and steroid exposure can be examined in future studies.

In conclusion, we performed a GWAS on patients with clinically and radiographically confirmed nontraumatic ONFH and population controls. We identified novel variants in regulatory genomic regions significantly associated with ONFH. Gene-based analyses identified PPARGC1B to be a significant and potentially functionally important gene for ONFH. Future confirmation in independent cohorts and with larger samples sizes is warranted.


Supporting material provided by the authors is posted with the online version of this article as a data supplement at (

Note: The authors thank the staff and participants of MyCode and the Regeneron Genetic Center for their support on data generation, processing, and storage; Jennifer L. Harding for project coordination; Dr. George F. Muschler for his advice; and Ilene Ladd for English editing.


1. Malizos KN, Karantanas AH, Varitimidis SE, Dailiana ZH, Bargiotas K, Maris T. Osteonecrosis of the femoral head: etiology, imaging and treatment. Eur J Radiol. 2007 Jul;63(1):16-28. Epub 2007 Jun 6.
2. Rajpura A, Wright AC, Board TN. Medical management of osteonecrosis of the hip: a review. Hip Int. 2011 Jul-Aug;21(4):385-92.
3. Mankin HJ. Nontraumatic necrosis of bone (osteonecrosis). N Engl J Med. 1992 May 28;326(22):1473-9.
4. Moya-Angeler J, Gianakos AL, Villa JC, Ni A, Lane JM. Current concepts on osteonecrosis of the femoral head. World J Orthop. 2015 Sep 18;6(8):590-601.
5. Mont MA, Hungerford DS. Non-traumatic avascular necrosis of the femoral head. J Bone Joint Surg Am. 1995 Mar;77(3):459-74.
6. Mont MA, Jones LC, Hungerford DS. Nontraumatic osteonecrosis of the femoral head: ten years later. J Bone Joint Surg Am. 2006 May;88(5):1117-32. Erratum in: J Bone Joint Surg Am. 2006 Jul;88(7):1602.
7. Karol SE, Yang W, Van Driest SL, Chang TY, Kaste S, Bowton E, Basford M, Bastarache L, Roden DM, Denny JC, Larsen E, Winick N, Carroll WL, Cheng C, Pei D, Fernandez CA, Liu C, Smith C, Loh ML, Raetz EA, Hunger SP, Scheet P, Jeha S, Pui CH, Evans WE, Devidas M, Mattano LA Jr, Relling MV. Genetics of glucocorticoid-associated osteonecrosis in children with acute lymphoblastic leukemia. Blood. 2015 Oct 8;126(15):1770-6. Epub 2015 Aug 11.
8. Karol SE, Mattano LA Jr, Yang W, Maloney KW, Smith C, Liu C, Ramsey LB, Fernandez CA, Chang TY, Neale G, Cheng C, Mardis E, Fulton R, Scheet P, San Lucas FA, Larsen EC, Loh ML, Raetz EA, Hunger SP, Devidas M, Relling MV. Genetic risk factors for the development of osteonecrosis in children under age 10 treated for acute lymphoblastic leukemia. Blood. 2016 Feb 4;127(5):558-64. Epub 2015 Nov 20.
9. Baek SH, Kim KI, Yoon KS, Kim TH, Kim SY. Genome-wide association scans for idiopathic osteonecrosis of the femoral head in a Korean population. Mol Med Rep. 2017 Feb;15(2):750-8. Epub 2016 Dec 14.
10. Sakamoto Y, Yamamoto T, Sugano N, Takahashi D, Watanabe T, Atsumi T, Nakamura J, Hasegawa Y, Akashi K, Narita I, Miyamoto T, Takeuchi T, Ikari K, Amano K, Fujie A, Kubo T, Tada Y, Kaneuji A, Nakamura H, Miyamura T, Kabata T, Yamaji K, Okawa T, Sudo A, Ohzono K, Tanaka Y, Yasunaga Y, Matsuda S, Imai Y, Akiyama M, Kubo M, Kamatani Y, Iwamoto Y, Ikegawa S; Japanese Research Committee on Idiopathic Osteonecrosis of the Femoral Head. Genome-wide association study of idiopathic osteonecrosis of the femoral Head. Sci Rep. 2017 Nov 8;7(1):15035.
11. Wang T, Azeddine B, Mah W, Harvey EJ, Rosenblatt D, Séguin C. Osteonecrosis of the femoral head: genetic basis. Int Orthop. 2019 Mar;43(3):519-30. Epub 2018 Oct 17.
12. Wyles CC, Paradise CR, Houdek MT, Slager SL, Terzic A, Behfar A, van Wijnen AJ, Sierra RJ. CORR® ORS Richard A. Brand Award: disruption in peroxisome proliferator-activated receptor-γ (PPARG) increases osteonecrosis risk through genetic variance and pharmacologic modulation. Clin Orthop Relat Res. 2019 Aug;477(8):1800-12.
13. Chang CC, Greenspan A, Gershwin ME. Osteonecrosis: current perspectives on pathogenesis and treatment. Semin Arthritis Rheum. 1993 Aug;23(1):47-69.
14. Carey DJ, Fetterolf SN, Davis FD, Faucett WA, Kirchner HL, Mirshahi U, Murray MF, Smelser DT, Gerhard GS, Ledbetter DH. The Geisinger MyCode community health initiative: an electronic health record-linked biobank for precision medicine research. Genet Med. 2016 Sep;18(9):906-13. Epub 2016 Feb 11.
15. Dewey FE, Murray MF, Overton JD, Habegger L, Leader JB, Fetterolf SN, O’Dushlaine C, Van Hout CV, Staples J, Gonzaga-Jauregui C, Metpally R, Pendergrass SA, Giovanni MA, Kirchner HL, Balasubramanian S, Abul-Husn NS, Hartzel DN, Lavage DR, Kost KA, Packer JS, Lopez AE, Penn J, Mukherjee S, Gosalia N, Kanagaraj M, Li AH, Mitnaul LJ, Adams LJ, Person TN, Praveen K, Marcketta A, Lebo MS, Austin-Tse CA, Mason-Suares HM, Bruse S, Mellis S, Phillips R, Stahl N, Murphy A, Economides A, Skelding KA, Still CD, Elmore JR, Borecki IB, Yancopoulos GD, Davis FD, Faucett WA, Gottesman O, Ritchie MD, Shuldiner AR, Reid JG, Ledbetter DH, Baras A, Carey DJ. Distribution and clinical impact of functional variants in 50,726 whole-exome sequences from the DiscovEHR study. Science. 2016 Dec 23;354(6319):aaf6814.
16. Zhang Y, Poler SM, Li J, Abedi V, Pendergrass SA, Williams MS, Lee MTM. Dissecting genetic factors affecting phenylephrine infusion rates during anesthesia: a genome-wide association study employing EHR data. BMC Med. 2019 Aug 28;17(1):168.
17. Chang CC, Chow CC, Tellier LC, Vattikuti S, Purcell SM, Lee JJ. Second-generation PLINK: rising to the challenge of larger and richer datasets. Gigascience. 2015 Feb 25;4:7.
18. de Leeuw CA, Mooij JM, Heskes T, Posthuma D. MAGMA: generalized gene-set analysis of GWAS data. PLoS Comput Biol. 2015 Apr 17;11(4):e1004219.
19. Zhang H, Zhang L, Wang J, Ma Y, Zhang J, Mo F, Zhang W, Yan S, Yang G, Lin B. Proteomic analysis of bone tissues of patients with osteonecrosis of the femoral head. OMICS. 2009 Dec;13(6):453-66.
20. Topaz O, Indelman M, Chefetz I, Geiger D, Metzker A, Altschuler Y, Choder M, Bercovich D, Uitto J, Bergman R, Richard G, Sprecher E. A deleterious mutation in SAMD9 causes normophosphatemic familial tumoral calcinosis. Am J Hum Genet. 2006 Oct;79(4):759-64. Epub 2006 Aug 24.
21. Hershkovitz D, Gross Y, Nahum S, Yehezkel S, Sarig O, Uitto J, Sprecher E. Functional characterization of SAMD9, a protein deficient in normophosphatemic familial tumoral calcinosis. J Invest Dermatol. 2011 Mar;131(3):662-9. Epub 2010 Dec 16.
22. Narumi S, Amano N, Ishii T, Katsumata N, Muroya K, Adachi M, Toyoshima K, Tanaka Y, Fukuzawa R, Miyako K, Kinjo S, Ohga S, Ihara K, Inoue H, Kinjo T, Hara T, Kohno M, Yamada S, Urano H, Kitagawa Y, Tsugawa K, Higa A, Miyawaki M, Okutani T, Kizaki Z, Hamada H, Kihara M, Shiga K, Yamaguchi T, Kenmochi M, Kitajima H, Fukami M, Shimizu A, Kudoh J, Shibata S, Okano H, Miyake N, Matsumoto N, Hasegawa T. SAMD9 mutations cause a novel multisystem disorder, MIRAGE syndrome, and are associated with loss of chromosome 7. Nat Genet. 2016 Jul;48(7):792-7. Epub 2016 May 16.
23. Davidsson J, Puschmann A, Tedgård U, Bryder D, Nilsson L, Cammenga J. SAMD9 and SAMD9L in inherited predisposition to ataxia, pancytopenia, and myeloid malignancies. Leukemia. 2018 May;32(5):1106-15. Epub 2018 Feb 25.
24. Narumi S, Hasegawa T. [Association between SAMD9/SAMD9L and hematological malignancies]. Rinsho Ketsueki. 2018;59(11):2475-80. Japanese.
25. Wong JC, Bryant V, Lamprecht T, Ma J, Walsh M, Schwartz J, Del Pilar Alzamora M, Mullighan CG, Loh ML, Ribeiro R, Downing JR, Carroll WL, Davis J, Gold S, Rogers PC, Israels S, Yanofsky R, Shannon K, Klco JM. Germline SAMD9 and SAMD9L mutations are associated with extensive genetic evolution and diverse hematologic outcomes. JCI Insight. 2018 Jul 26;3(14):121086.
26. Lopes da Silva M, O’Connor MN, Kriston-Vizi J, White IJ, Al-Shawi R, Simons JP, Mössinger J, Haucke V, Cutler DF. Type II PI4-kinases control Weibel-Palade body biogenesis and von Willebrand factor structure in human endothelial cells. J Cell Sci. 2016 May 15;129(10):2096-105. Epub 2016 Apr 11.
27. Bijnens AP, Gils A, Jutten B, Faber BC, Heeneman S, Kitslaar PJ, Tordoir JH, de Vries CJ, Kroon AA, Daemen MJ, Cleutjens KB. Vasculin, a novel vascular protein differentially expressed in human atherogenesis. Blood. 2003 Oct 15;102(8):2803-10. Epub 2003 Jul 3.
28. Andersen G, Wegner L, Yanagisawa K, Rose CS, Lin J, Glümer C, Drivsholm T, Borch-Johnsen K, Jørgensen T, Hansen T, Spiegelman BM, Pedersen O. Evidence of an association between genetic variation of the coactivator PGC-1beta and obesity. J Med Genet. 2005 May;42(5):402-7.
29. Delgado-Lista J, Perez-Martinez P, Solivera J, Garcia-Rios A, Perez-Caballero AI, Lovegrove JA, Drevon CA, Defoort C, Blaak EE, Dembinska-Kieć A, Risérus U, Herruzo-Gomez E, Camargo A, Ordovas JM, Roche H, Lopez-Miranda J. Top single nucleotide polymorphisms affecting carbohydrate metabolism in metabolic syndrome: from the LIPGENE study. J Clin Endocrinol Metab. 2014 Feb;99(2):E384-9. Epub 2013 Nov 7.
30. Franks PW, Christophi CA, Jablonski KA, Billings LK, Delahanty LM, Horton ES, Knowler WC, Florez JC; Diabetes Prevention Program Research Group. Common variation at PPARGC1A/B and change in body composition and metabolic traits following preventive interventions: the Diabetes Prevention Program. Diabetologia. 2014 Mar;57(3):485-90. Epub 2013 Dec 7.
31. Villegas R, Williams SM, Gao YT, Long J, Shi J, Cai H, Li H, Chen CC, Tai ES, Hu F, Cai Q, Zheng W, Shu XO; AGEN-T2D Consortium. Genetic variation in the peroxisome proliferator-activated receptor (PPAR) and peroxisome proliferator-activated receptor gamma co-activator 1 (PGC1) gene families and type 2 diabetes. Ann Hum Genet. 2014 Jan;78(1):23-32.
32. Zhang Y, Rohatgi N, Veis DJ, Schilling J, Teitelbaum SL, Zou W. PGC1β organizes the osteoclast cytoskeleton by mitochondrial biogenesis and activation. J Bone Miner Res. 2018 Jun;33(6):1114-25. Epub 2018 Mar 11.
33. Ju D, He J, Zhao L, Zheng X, Yang G. Estrogen related receptor α-induced adipogenesis is PGC-1β-dependent. Mol Biol Rep. 2012 Mar;39(3):3343-54. Epub 2011 Jul 6.
34. Chang HS, Lee SH, Lee JU, Park JS, Chung IY, Park CS. Functional characterization of exonic variants of the PPARGC1B gene in coregulation of estrogen receptor alpha. DNA Cell Biol. 2016 Jul;35(7):314-21. Epub 2016 Mar 30.
35. Kang JS, Park S, Song JH, Jung YY, Cho MR, Rhyu KH. Prevalence of osteonecrosis of the femoral head: a nationwide epidemiologic analysis in Korea. J Arthroplasty. 2009 Dec;24(8):1178-83. Epub 2009 Jul 28.
36. Fukushima W, Fujioka M, Kubo T, Tamakoshi A, Nagai M, Hirota Y. Nationwide epidemiologic survey of idiopathic osteonecrosis of the femoral head. Clin Orthop Relat Res. 2010 Oct;468(10):2715-24. Epub 2010 Mar 12.
37. Liu F, Wang W, Yang L, Wang B, Wang J, Chai W, Zhao D. An epidemiological study of etiology and clinical characteristics in patients with nontraumatic osteonecrosis of the femoral head. J Res Med Sci. 2017 Feb 16;22:15.
38. Bleker SM, Coppens M, Middeldorp S. Sex, thrombosis and inherited thrombophilia. Blood Rev. 2014 May;28(3):123-33. Epub 2014 Apr 1.
39. Del Principe D, Ruggieri A, Pietraforte D, Villani A, Vitale C, Straface E, Malorni W. The relevance of estrogen/estrogen receptor system on the gender difference in cardiovascular risk. Int J Cardiol. 2015;187:291-8. Epub 2015 Mar 17.
40. McCarthy NS, Vangjeli C, Surendran P, Treumann A, Rooney C, Ho E, Sever P, Thom S, Hughes AD, Munroe PB, Howard P, Johnson T, Caulfield M, Shields DC, O’Brien E, Fitzgerald DJ, Stanton AV. Genetic variants in PPARGC1B and CNTN4 are associated with thromboxane A2 formation and with cardiovascular event free survival in the Anglo-Scandinavian Cardiac Outcomes Trial (ASCOT). Atherosclerosis. 2018 Feb;269:42-9. Epub 2017 Dec 8.

Supplemental Digital Content

Copyright © 2020 The Authors. Published by The Journal of Bone and Joint Surgery, Incorporated All rights reserved.