Risk factors for fractures and avascular osteonecrosis in type 1 Gaucher disease: A study from the International Collaborative Gaucher Group (ICGG) Gaucher Registry
We hypothesized that overall disease activity or the severity of involvement of individual disease compartments, as measured by clinical and surrogate markers, predict the risk of avascular osteonecrosis (AVN) or fractures in type 1 Gaucher disease (GD1). We applied our risk-set matched case-control method to identify four patient groups within the International Collaborative Gaucher Group (ICGG) Gaucher Registry based on the presence and absence of AVN and fractures. Characteristics of GD1 were examined by comparing the distributions of each risk factor in cases versus matched controls using conditional logistic regression to calculate adjusted odds ratios (OR). Potential risk factors included hematological and visceral parameters, GD1 biomarkers, white blood cells, GBA1 genotype, and spine and femur dual-energy X-ray absorptiometry (DXA) Z-scores. In the total population of 5894 ICGG Gaucher Registry patients, 544 experienced at least one episode of AVN; 2008 reported no history of AVN. Clinical and surrogate markers of disease activity were similar in patients with and without AVN; patients with AVN were 1.6 times more likely to be anemic compared to matched controls (OR = 1.59; 95% confidence interval [CI], 1.06–2.38, p < 0.05). For fractures, 319 patients suffered fractures and 1233 had no prior history of fractures. Clinical and surrogate markers of disease in patients with and without fractures were similar, except for mean lumbar spine DXA Z-scores. Among patients with fractures, 49.3% had DXA Z-scores ≤ −1 compared to 31.0% in the control group. Compared to controls with Z-scores > −1.0, GD1 patients exhibiting Z-scores ≤ −1 had an OR of 5.55 (95% CI, 1.81–17.02, p < 0.01) for fracture. In GD1, after controlling for gender, year of birth, treatment status, and splenectomy status, we identified new risk factors for AVN and fractures. Concurrent anemia was associated with an increased risk for AVN. Low bone mineral density of the lumbar spine was a strong risk factor for fractures of the spine and femur in GD1. © 2012 American Society for Bone and Mineral Research.
Gaucher disease (GD) is the most common lysosomal storage disorder. It results from autosomal recessive mutations in the GBA1 gene that encodes acid β-glucosidase (EC 184.108.40.206; lysosomal glucocerebrosidase). Deficient acid β-glucosidase activity leads to the accumulation of glucocerebroside in the lysosomes of mononuclear phagocytes and a complex multisystemic phenotype.1 Type 1 GD (GD1) comprises the majority (∼94%) of currently known cases, and it is differentiated from type 2 and type 3 GD by the absence of neurodegenerative disease.1 The conspicuous sites of pathology in GD1 are the liver, the spleen, the bone marrow, and the skeleton,1, 2 although considerable phenotypic heterogeneity exists between patients, even among those with the same GBA1 genotype.3, 4
Clinical and radiological evidence of skeletal involvement is found in the vast majority of patients.1, 2, 5 Although clinical observations indicate that severe skeletal manifestations may occur in the absence of commensurate severe visceral and hematological disease, it is not known whether, in general, there is a relationship between the severity of skeletal disease and the severity of visceral/hematological disease. This question would be illuminated by a better understanding of the risk factors for avascular osteonecrosis (AVN) and fractures in GD1.
Bone involvement in GD1 is multidimensional and essentially every compartment of this organ is involved. Compared to the visceral and the hematological disease, bone involvement in patients with GD1 causes the greatest impairment in the quality of life.6 The most impactful and dramatic clinical complications arising from bone involvement are AVN, osteopenia, and fractures.2, 5, 7, 8 These complications permanently alter the health status of affected patients due to chronic pain, disability, and the need for surgical interventions.6 The precise risk of AVN in untreated GD1 (not receiving imiglucerase therapy) was estimated to be 22.8 per 1000 person years (95% confidence interval [CI], 20.2–25.7) of follow-up.9 Imiglucerase therapy reduces the incidence of AVN to 13.8 per 1000 years of follow-up.10 Maximal reduction in the risk of AVN upon treatment with imiglucerase occurs among patients who begin treatment within 2 years of diagnosis of GD1.10 The pathophysiological mechanisms underlying AVN in GD1 are not understood. Other than the most conspicuous form of AVN, akin to that seen in sickle cell crises, there is also a silent variety of AVN in GD1 that occurs in the form of medullary infarction due to asymptomatic vascular obstruction of medullary blood vessels; this form of AVN is detected by magnetic resonance imaging (MRI) of the bone marrow.11
The majority of GD1 patients exhibit osteopenia,2, 5, 7 measured by dual-energy X-ray absorptiometry (DXA).12, 13 Although osteopenia in GD is commonly attributed to increased osteoclastic bone resorption,14 it is more likely related to impaired osteoblast function15, 16 associated with accumulating lipids.17, 18 Importantly, it is not known whether osteopenia in GD1 increases the risk of fractures as it does in postmenopausal women.19 Imiglucerase therapy improves bone mass and bone mineral density12, 13, 20–23 and studies have shown an apparent decrease in reports of bone pain and bone crises.23, 24 However, it is not known whether imiglucerase treatment decreases the risk of fractures.
The aims of our study were to identify risk factors for AVN and fractures in GD1. We hypothesized that the presence of increased disease activity, as measured by clinical and surrogate markers, would predict the risk of AVN or fractures in GD1. The study was performed using the International Collaborative Gaucher Group (ICGG) Gaucher Registry database of currently over 6000 GD patients, most of whom have GD1. Here, we report the risk factors identified for both AVN and fractures.
Patients and Methods
ICGG Gaucher Registry
The ICGG Gaucher Registry has been in existence since 1991 (ClinicalTrials.gov NCT003589430). It is a global registry of the clinical, demographic, genetic, biochemical, and treatment characteristics of GD patients. It represents one of the largest registries for a rare disease. The key objectives of the Registry are to define the clinical spectrum of GD, assess its natural history though longitudinal follow-up, and assess the effect of treatment. An independent international group of physician experts in GD provides scientific direction and governance of the Registry, with logistical support from Genzyme, a Sanofi company (Cambridge, MA, USA). More than 700 physicians from some 60 countries participate in the Registry and, following local Institutional Review Board approvals, over 6000 patients have been enrolled in the ICGG Gaucher Registry.
Study design, population, and case-control matching
Due to the stochastic nature of both AVN and fractures in GD, we used a case-control study design to delineate risk factors for these complications. To reduce the chance of biased selection of cases and controls, we employed our recently described risk-set method for case-control studies in rare disease registries.25
We identified all patients in the ICGG Gaucher Registry as of October 1, 2010 who met the following inclusion criteria: confirmed diagnosis of GD1, known treatment status, and known date of initiation of imiglucerase (Cerezyme, Genzyme) or alglucerase (Ceredase, Genzyme) treatment. Until 2010, alglucerase and imiglucerase were the only commercially approved enzyme treatments for GD1. Alglucerase and imiglucerase have been shown to be therapeutically equivalent in a randomized clinical trial.26 Therefore, these two treatments will be denoted as imiglucerase in this publication.
Cases of AVN and fractures were identified based on affirmative radiographic or MRI reports in the ICGG Registry. For each potential case patient, we reviewed skeletal history and individual skeletal assessments reported to the Registry to identify the earliest date of either AVN or fractures. Similarly, we identified all patients who had not suffered AVN or fractures, and these patients served as controls after appropriate matching. Matched case-control groups were formed for AVN and for fractures, according to gender, year of birth (± 5 years), treatment status, and splenectomy status at the time of the index date.25 Patients were matched for spleen status because splenectomy is a recognized risk factor for AVN.8, 10
Risk factors for AVN or fracture were determined following matching, as described in the previous section. GBA1 genotype and phenotype pertaining to hematological and visceral compartments were analyzed in the cases and the controls. Additionally, biomarkers were examined as indicators of overall disease activity including chitotriosidase, angiotensin-converting enzyme (ACE), tartrate-resistant acid phosphatase (TRAP), and peripheral white blood cell (WBC) count. A major focus of the study was to determine whether osteopenia, as indicated by lumbar spine and femur DXA Z-scores, was a risk factor for fractures in GD1. For each candidate risk factor, we identified the closest available assessment that occurred no more than 2 years prior to the onset of the event among all cases and the matched index date among control patients. All laboratory results and DXA scores were reported to the Registry by the participating clinicians.
Frequency distributions for each candidate risk factor were constructed for the cases and the control patients. Presence of anemia was determined based on age and gender norms for hemoglobin (g/dL) concentrations as follows: <12 g/dL for males older than 12 years; <11 g/dL for females older than 12 years; <10.5 g/dL for children aged >2 to 12 years; <9.5 g/dL for children aged 6 months to 2 years; and <10.1 g/dL for children younger than 6 months of age. Thrombocytopenia was defined as platelet count <120 × 103/mm3. Liver and spleen volumes were determined by volumetric imaging and the results were expressed as multiples of normal (MN) volume, based on body weight; eg, normal liver volume is 2.5% of body weight and normal spleen volume is 0.2% of body weight.
We determined the 75th percentile for the biomarkers chitotriosidase, ACE, and TRAP due to variability in choice of substrate, laboratory technique, and lack of standardized clinical norms. A dichotomous indicator was created for cases and controls: patients with values ≤75th percentile and patients with values >75th percentile. Using this method, in cases and controls, we compared the percentages of patients who have the highest biomarker values, as the magnitude of the biomarker chitotriosidase elevation correlates with disease severity.27, 28 We were interested in analyzing WBC count because the severity of leukopenia in GD1 patients tends to parallel GD1 severity. We performed a similar analysis, categorizing patients into those with a WBC count <25th percentile and those with a WBC count ≥25th percentile. For the analysis of spine and femur DXA Z-scores, we used cutoffs of ≤ −1 versus > −1. In patients with fractures, additional analysis was performed with respect to the site of fractures and median age at the time of fracture.
In addition to constructing frequency distributions for each candidate risk factor, we used conditional logistic regression to produce adjusted odds ratios (OR), 95% CIs, and p values for each candidate risk factor, controlling for the four matching characteristics (gender, year of birth, treatment status, and splenectomy status).25 An adjusted OR of 1.0 indicates no difference in the distributions between cases and controls. For GBA1 genotype, a Cochran-Mantel-Haenszel test of general association, controlling for case-control matching, was used to determine whether significant differences in the distributions for cases and controls existed. All tests were performed at the alpha = 0.05 level. All analyses were conducted using SAS 9.2 (SAS Institute, Inc., Cary, NC, USA) in accordance with Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines.
From the population of 5894 GD patients enrolled in the ICGG Gaucher Registry as of October 1, 2010, a total of 5156 patients met the inclusion criteria of this study. From this group of patients (n = 5156), 176 patients had a history of AVN and 416 patients a history of fractures with no accompanying assessment or diagnosis dates reported to the Registry; they were therefore excluded from the study. Of the remaining patients, we identified and matched 544 patients with reports of AVN with 2008 patients without AVN, and we matched 319 patients with reports of fractures with 1233 patients without fractures.
The characteristics used for case-control matching in the AVN groups are shown in Table 1. Following matching, the ratio of females to males was similar in both groups, as were the distributions of patients born in each decade.
Table 1. Characteristics Used in Case-Control Matching of Patients With and Without Avascular Osteonecrosis
|Gender, n (%)|
| Males||258 (47.4)||960 (47.8)|
| Females||286 (52.6)||1048 (52.2)|
|Year of birth, n (%)|
| 1910–<1920||3 (0.6)||17 (0.8)|
| 1920–<1930||14 (2.6)||66 (3.3)|
| 1930–<1940||57 (10.5)||148 (7.4)|
| 1940–<1950||71 (13.1)||257 (12.8)|
| 1950–<1960||90 (16.5)||315 (15.7)|
| 1960–<1970||70 (12.9)||303 (15.1)|
| 1970–<1980||103 (18.9)||342 (17.0)|
| 1980–<1990||88 (16.2)||281 (14.0)|
| 1990–<2000||45 (8.3)||247 (12.3)|
| 2000–<2010||3 (0.6)||32 (1.6)|
|Treatment status, n (%)|
| Untreated at the time of the index date||317 (58.3)||1184 (59.0)|
| Treated at the time of the index date||227 (41.7)||824 (41.0)|
|Splenectomy status, n (%)|
| Spleen intact at the time of the index date||375 (68.9)||1531 (76.2)|
| Splenectomized at the time of the index date||169 (31.1)||477 (23.8)|
GBA1 gene mutations were reported for over 70% of patients with and without AVN (Table 2). The majority of patients had at least one N370S allele, with an excess of N370S homozygosity in patients who had not suffered AVN (with AVN: 23.8% homozygous, 63.5% heteroallelic; without AVN: 36.5% homozygous, 52.8% heteroallelic). In N370S heteroallelic patients, the excess risk in the AVN group appeared to be confined to the severe mutations 84GG, L444P or Other (Table 2). A Cochran-Mantel-Haenszel test of general association resulted in statistically significant (p = 0.0018) differences in the distribution of genotypes between patients with and without AVN. Detailed GBA1 genotype information is presented in Supplementary Table 1.
Table 2. GBA1 Genotypes of Patients With and Without Avascular Osteonecrosis
|N370S/N370S||102 (23.8)||525 (36.5)|
|N370S/L444P||90 (21.0)||233 (16.2)|
|N370S/84GG||44 (10.3)||108 (7.5)|
|N370S/otherb||138 (32.2)||418 (29.1)|
|All others||55 (12.8)||153 (10.6)|
The clinical characteristics of patients with and without AVN are depicted in Table 3. The patients in the two groups were similar in most clinical parameters: presence of hepatomegaly and splenomegaly, platelet count, biomarkers, white cell count, and DXA Z-score. The exception was anemia; patients with anemia were 60% more likely to develop AVN compared to patients without anemia (OR, 1.59; 95% CI, 1.06–2.38; p < 0.05). However, patients with and without AVN were not distinguishable with respect to the severity of anemia. Mean hemoglobin concentrations of anemic patients were similar in the two groups (mean ± SD): with AVN 10.0 ± 1.12 g/dL, n = 80; without AVN 9.8 ± 1.35 g/dL, n = 118. The mean hemoglobin concentrations of patients without anemia were also very similar: with AVN 13.2 ± 1.42 g/dL, n = 292; without AVN: 13.4 ± 1.39 g/dL, n = 872).
Table 3. Hematological, Visceral, Biomarkers, and Skeletal Manifestations Among Patients With and Without Avascular Osteonecrosis
|Total number of patients||544||2008|| || || |
|Anemia status, n (%)a||n = 372||n = 990|| || || |
| No||292 (78.5)||872 (88.1)||Reference|| || |
| Yes||80 (21.5)||118 (11.9)||1.59||(1.06, 2.38)||0.0249|
|Thrombocytopenia status, n (%), (platelet count, ×103/mm3)||n = 371||n = 1005|| || || |
| ≥120||214 (57.7)||633 (63.0)||Reference|| || |
| <120||157 (42.3)||372 (37.0)||1.08||(0.97, 1.21)||0.1670|
|Hepatomegaly status, n (%), (liver volume in multiples of normal)||n = 177||n = 448|| || || |
| ≤1.25||91 (51.4)||309 (69.0)||Reference|| || |
| >1.25||86 (48.6)||139 (31.0)||1.03||(0.83, 1.28)||0.7720|
|Splenomegaly status, n (%), (spleen volume in multiples of normal)||n = 141||n = 426|| || || |
| ≤5||41 (29.1)||144 (33.8)||Reference|| || |
| >5||100 (70.9)||282 (66.2)||1.00||(0.83, 1.22)||0.9647|
|Chitotriosidase, n (%) (nmol/mL/hr), (75th percentile = 5219.50)||n = 107||n = 357|| || || |
| ≤75th percentile||73 (68.2)||275 (77.0)||Reference|| || |
| >75th percentile||34 (31.8)||82 (23.0)||1.50||(0.68, 3.29)||0.3153|
|Angiotensin converting enzyme, n (%), (U/L), (75th percentile = 137.00)||n = 132||n = 327|| || || |
| ≤75th Percentile||84 (63.6)||261 (79.8)||Reference|| || |
| >75th Percentile||48 (36.4)||66 (20.2)||1.69||(0.60, 4.73)||0.3162|
|White blood cell count, n (%), (×103/mm3), (25th percentile = 4.60)||n = 120||n = 283|| || || |
| ≥25th Percentile||94 (78.3)||210 (74.2)||Reference|| || |
| <25th Percentile||26 (21.7)||73 (25.8)||0.86||(0.37, 1.99)||0.7200|
|Tartrate resistant acid phosphatase, n (%), (U/L), (75th percentile = 12.80)||n = 202||n = 535|| || || |
| ≤75th Percentile||137 (67.8)||417 (77.9)||Reference|| || |
| >75th Percentile||65 (32.2)||118 (22.1)||1.59||(0.93, 2.71)||0.0911|
|Spine lumbar DXA Z-score, n (%)||n = 77||n = 222|| || || |
| Z-score > −1||43 (55.8)||137 (61.7)||Reference|| || |
| Z-score ≤ −1||34 (44.2)||85 (38.3)||0.82||(0.37, 1.82)||0.6326|
|Femur DXA Z-score, n (%)||n = 47||n = 145|| || || |
| Z-score > −1||31 (66.0)||97 (66.9)||Reference|| || |
| Z-score ≤ −1||16 (34.0)||48 (33.1)||1.12||(0.22, 5.62)||0.8906|
The characteristics used for matching cases and controls for fractures are shown in Table 4. No statistical differences were observed between the cases and controls.
Table 4. Characteristics Used in Case-Control Matching of Patients With and Without Fractures
|Gender, n (%)|
| Males||149 (46.7)||613 (49.7)|
| Females||170 (53.3)||620 (50.3)|
|Year of birth, n (%)|
| 1910–<1920||5 (1.6)||17 (1.4)|
| 1920–<1930||18 (5.6)||61 (4.9)|
| 1930–<1940||37 (11.6)||100 (8.1)|
| 1940–<1950||54 (16.9)||192 (15.6)|
| 1950–<1960||51 (16.0)||203 (16.5)|
| 1960–<1970||47 (14.7)||194 (15.7)|
| 1970–<1980||40 (12.5)||177 (14.4)|
| 1980–<1990||41 (12.9)||171 (13.9)|
| 1990–<2000||21 (6.6)||105 (8.5)|
| 2000–<2010||5 (1.6)||13 (1.1)|
|Treatment status, n (%)|
| Untreated at the time of the index date||135 (42.3)||539 (43.7)|
| Treated at the time of the index date||184 (57.7)||694 (56.3)|
|Splenectomy status, n (%)|
| Spleen intact at the time of the index date||209 (65.5)||871 (70.6)|
| Splenectomized at the time of the index date||110 (34.5)||362 (29.4)|
GBA1 gene mutations were reported for the majority of patients with and without fractures (Table 5). Most patients had at least one N370S allele (with fractures: 31.6% homozygous, 55.7% heteroallelic; without fractures: 34.7% homozygous, 53.2% heteroallelic). Unlike the situation with AVN, there was no difference (p = 0.8519) in GBA1 genotype distributions between the two groups of patients with and without fractures. Detailed GBA1 genotype information is presented in Supplementary Table 2.
Table 5. GBA1 Genotypes of Patients With and Without Fractures
|N370S/N370S||84 (31.6)||315 (34.7)|
|N370S/L444P||42 (15.8)||158 (17.4)|
|N370S/84GG||21 (7.9)||81 (8.9)|
|N370S/othera||85 (32.0)||244 (26.9)|
|All others||34 (12.8)||110 (12.1)|
The clinical characteristics of patients with and without fractures are depicted in Table 6. The patients in the two groups were similar in most clinical parameters with the exception of the distribution of lumbar spine DXA Z-scores. Patients with DXA Z-scores ≤ −1 had an OR of 5.55 (p < 0.01), indicating a more than fivefold increase in risk to experience a fracture compared to patients with Z-scores > −1. As a group, patients with mean DXA Z-scores ≤ −1 was −1.8 (with fractures: −1.8 ± 0.77; without fractures: −1.8 ± 0.65). The average Z-scores of the groups with Z-scores > −1 for patients with and without fractures were also similar (with fractures: 0.5 ± 1.23; without fractures: 0.4 ± 0.92).
Table 6. Hematological, Visceral, Biomarkers, and Skeletal Manifestations Among Patients With and Without Fractures
|Total number of patients||319||1233|| || || |
|Anemia status, n (%)a||n = 238||n = 710|| || || |
| No||196 (82.4)||628 (88.5)||Reference|| || |
| Yes||42 (17.6)||82 (11.5)||1.56||(0.96, 2.52)||0.0711|
|Thrombocytopenia status, n (%), (platelet count, ×103/mm3)||n = 237||n = 724|| || || |
| ≥120||150 (63.3)||491 (67.8)||Reference|| || |
| <120||87 (36.7)||233 (32.2)||1.08||(0.94, 1.23)||0.2919|
|Hepatomegaly status, n (%), (liver volume in multiples of normal)||n = 110||n = 339|| || || |
| ≤1.25||71 (64.5)||230 (67.8)||Reference|| || |
| >1.25||39 (35.5)||109 (32.2)||1.14||(0.91, 1.43)||0.2438|
|Splenomegaly status, n (%), (spleen volume in multiples of normal)||n = 94||n = 284|| || || |
| ≤5||45 (47.9)||106 (37.3)||Reference|| || |
| >5||49 (52.1)||178 (62.7)||0.90||(0.72, 1.11)||0.3038|
|Chitotriosidase, n (%), (nmol/mL/hr), (75th percentile = 4315.50)||n = 79||n = 273|| || || |
| ≤75th percentile||52 (65.8)||212 (77.7)||Reference|| || |
| >75th percentile||27 (34.2)||61 (22.3)||1.48||(0.69, 3.21)||0.3172|
|Angiotensin converting enzyme, n (%), (U/L), (75th percentile = 114.00)||n = 85||n = 282|| || || |
| ≤75th percentile||58 (68.2)||218 (77.3)||Reference|| || |
| >75th percentile||27 (31.8)||64 (22.7)||1.07||(0.38, 3.04)||0.9001|
|White blood cell count, n (%), (×103/mm3), (25th percentile = 4.60)||n = 80||n = 211|| || || |
| ≥25th percentile||60 (75.0)||162 (76.8)||Reference|| || |
| <25th percentile||20 (25.0)||49 (23.2)||1.10||(0.46, 2.63)||0.8241|
|Tartrate resistant acid phosphatase, n (%), (U/L), (75th percentile = 11.80)||n = 135||n = 398|| || || |
| ≤75th percentile||89 (65.9)||313 (78.6)||Reference|| || |
| >75th percentile||46 (34.1)||85 (21.4)||1.49||(0.82, 2.69)||0.1908|
|Spine lumbar DXA Z-score, n (%)||n = 71||n = 171|| || || |
| Z-score > −1||36 (50.7)||118 (69.0)||Reference|| || |
| Z-score ≤ −1||35 (49.3)||53 (31.0)||5.55||(1.81, 17.02)||0.0027|
|Femur DXA Z-score, n (%)||n = 739||n = 7127|| || || |
| Z-score > −1||30 (76.9)||86 (67.7)||Reference|| || |
| Z-score ≤ −1||9 (23.1)||41 (32.3)||1.63||(0.36, 7.43)||0.5259|
The ICGG Gaucher Registry database also includes the site of reported fractures. Among patients included in this analysis, sites were reported for 327 fractures (Table 7). The most common site of fracture in GD1 patients was the spine, which accounted for 36.4% of all fractures (119/327), followed by the lower extremities (distal to the knee, femur, hip, knee), where 34.9% of fractures occurred. The fractures occurred at a relatively young age: the median age at fracture for these sites were 45.9 years, 38.2 years, 33.8 years, 56.9 years, and 34.1 years, respectively.
Table 7. Sites of Fractures in All Patients Reporting Fractures
|Spine||119 (36.4)||45.9 (32.4, 60.3)|
|Distal to the knee||44 (13.5)||38.2 (17.4, 48.3)|
|Femur||43 (13.2)||33.8 (14.6, 45.6)|
|Distal to the elbow||32 (9.8)||29.3 (12.9, 51.0)|
|Rib||30 (9.2)||51.3 (40.3, 57.9)|
|Hip||21 (6.4)||56.9 (27.7, 62.4)|
|Humerus||18 (5.5)||46.2 (33.8, 61.8)|
|Clavicle||10 (3.1)||33.3 (16.6, 44.4)|
|Knee||6 (1.8)||34.1 (22.9, 54.1)|
|Elbow||3 (0.9)||52.1 (22.9, 71.2)|
As a result of the heterogeneous phenotype of GD1 and its highly variable natural history, it has not been possible to predict the occurrence of major life-altering skeletal complications, AVN, and fractures. A better understanding of the context in which these complications occur would help improve patient management through appropriate monitoring and directed therapy of high-risk patients. Therefore, in this study we aimed to define the risk factors for AVN and fractures in GD1. A priori, occurrence of such major complications would be expected to correlate with overall severity of GD1, as indicated by visceral parameters and biomarkers. Surprisingly, the burden of visceral disease, indicated by either the extent of hepatomegaly and splenomegaly, or the total body burden of Gaucher cells denoted by serum biomarkers, appeared not to predict the occurrence of AVN or fractures. Our results identified anemia as the main predictor of AVN, and osteopenia as a strong risk factor for fractures in GD1.
Importantly, this study is the first demonstration that low bone density (Z-score < −1.0) in GD1 is associated with a high risk for fractures (OR, 5.55). In GD, fractures result in a major negative impact on quality of life and could trigger a sequence of events leading to life-threatening complications, as hip fractures do in the general population.29 The identification of osteopenia as a risk factor for fractures in GD1 is important for the management of patients because osteopenia is highly prevalent in symptomatic GD1 patients, as well as asymptomatic N370S homozygous patients diagnosed through genetic screening programs.3, 4, 30, 31 We found that fractures occur in patients with GD1 in the range of relatively young median ages, from 29 to 57 years, as well as in older patients.
This result should be evaluated with the knowledge that, in the general population, for every 1 SD decrease in T-score, there is a 1.5-fold increase in fracture risk.32 In our population, we did not discriminate the age at which the fracture occurred, and thus Z-scores were used (as is customary in patients under 18 years of age) for all patients. In this regard, the Z-score would estimate a lower fracture risk than the T-score. With this in mind, the OR of 5.55 for a Z-score of ≤ −1 suggests a much higher fracture risk for this change in Z-score compared to a non-Gaucher population. Thus, the most significant determinant of fracture risk is the DXA T-score or Z-score, and amelioration of osteopenia should be a key therapeutic goal in the treatment of GD1.33
The results of our study indicate that, in the management of patients with GD1, a spinal DXA Z-score < −1 should be a significant trigger for therapeutic intervention directed at maintaining bone mineral density above this value. Macrophage-directed imiglucerase therapy improves bone density,12, 23 with striking responses in children and young adults.13
The addition of oral alendronate (40 mg/d) to imiglucerase-treated GD1 patients substantially increased the rapidity and magnitude of the bone mineral density response in premenopausal women and in men to age 70 years.34 However, a benefit in terms of fracture outcome was not investigated within the time frame of the study, and a role for bisphosphonate therapy for GD1-related osteopenia/osteoporosis, although commonly prescribed, remains conjectural. In an uncontrolled case study, the 10-year probability of a new fracture in elderly GD1 women treated with imiglucerase was twofold to threefold greater than that expected in comparably aged osteoporotic women without GD.35 There was no difference in bisphosphonate use between GD1 patients with or without new fractures, and DXA T-scores generally remained abnormal despite bisphosphonate use.35
In our study, although we cannot exclude some underreporting, it seems unlikely that the use of concurrent bisphosphonates had an impact on the analysis of either AVN or fractures. Of the 544 patients with AVN, only 2.9% (n = 16) were reported to have received bisphosphonates prior to the event, a percentage essentially indistinguishable from the 3.2% (65/2008) of patients without AVN who were reported as having received bisphosphonates prior to the index date. Of the 319 patients with fractures, 25 (7.8%) had reported bisphosphonate use prior to the fracture event compared with 4.0% (49/1233) prior to the index date in the control group without fractures.
Given the role of osteoblast dysfunction in the development of osteopenia of GD1,11, 17, 18 it will be important to determine whether alternative treatments with small molecule substrate reduction therapy, whose volume of distribution includes cells other than macrophages, result in an enhanced therapeutic effect. In a pooled retrospective study of GD1 patients, substrate reduction therapy with miglustat appeared to modestly increase bone mineral density.36 Eliglustat, a more potent and specific inhibitor of glucosylceramide synthase that is currently in clinical trials for GD1, was reported to increase lumbar spine bone mineral density after 1 and 2 years of treatment, with major gains in DXA scores in osteopenic and osteoporotic patients.37, 38
The signature complication of GD1 in the skeleton is the occurrence of AVN, which is clinically devastating when it involves the cortical bone. As with fractures, we did not find any association of AVN with the severity of visceral disease or with biomarkers. Although AVN is a classic manifestation of GD, other etiologies should also be kept in mind when evaluating patients.10, 11, 39 Other than the well-known association of splenectomy with AVN,8, 10 the only risk factor that emerged for AVN was anemia (OR, 1.59). Although GD1 patients with AVN were 1.6 times more likely to be anemic compared to patients without AVN, the two groups were not distinguishable with respect to the severity of anemia. It should be kept in mind that the patients were classified as anemic or nonanemic based on a single hemoglobin measurement in proximity to the date of AVN, or to the matched index date in the case of the non-AVN control group, and therefore it is not possible to determine whether the burden of anemia over time was associated with a greater frequency of AVN as would logically be expected if anemia itself was causative.
An alternative explanation for this novel association is that, in some patients, anemia may not reflect overall GD1 severity, but rather represent a biomarker of bone and bone marrow pathologies that are directly or indirectly associated with AVN. For example, a decrease in the mesenchymal stem-cell pool in the proximal femur was found in corticosteroid-induced osteonecrosis.40 Interestingly, increased osteoblast apoptosis and resulting loss of vascular endothelial growth factor–dependent angiogenesis support has been implicated in humans and in animal models of AVN receiving glucocorticoids.41–43 Osteoblastic dysfunction17 and abnormalities of mesenchymal stem cells44 have been described in animal models of GD and in a human patient, respectively. Both these cell types are essential components of the hematopoietic microenvironment, and dysfunction of either or both could conceivably link AVN and anemia as secondary phenomena. Thus it seems likely that, aside from a role in osteopenia, osteoblast dysfunction mediated via lipids that accumulate in GD may also contribute to AVN in GD. Although imiglucerase therapy markedly reduces the risk of AVN, especially among patients who begin treatment within 2 years of diagnosis,10 it will be of interest to determine whether small-molecule substrate reduction therapy via a potent inhibitor, such as eliglustat tartrate,45 will further minimize the risk of AVN through its broader cellular distribution.
There are other potential contributors to AVN in GD. For example, osteoblast and mesenchymal stromal cells as well as bone marrow adipocytes (negative regulators of hematopoiesis) participate in and are modulated by the systemic regulation of energy metabolism.46 Glycosphingolipid accumulation has been associated with insulin resistance.47 In GD1, glycosphingolipid accumulation is also associated with abnormal cholesterol metabolism48 and non–obesity dependent decreases in serum adiponectin levels.49 It is hypothesized by Youm and colleagues43 that abnormalities in blood cholesterol and triglyceride levels, sometimes associated with alcohol consumption, may be associated with nontraumatic osteonecrosis of the femoral head. Because information on these metabolic variables is lacking in the data collected by the Gaucher Registry, we focused on anemia as a common factor and found it to be a major risk factor for GD1-associated AVN. Therefore, anemia is an indicator of the interaction between erythroid development and skeletal homeostasis.50
In GD1, bone infarction by infiltration of Gaucher cells has been proposed as a mechanism, but all of our other clinical and surrogate markers showed no difference in disease severity of treatment status between patients with and without AVN. Most hypotheses in the pathogenesis of AVN involve the reduction of blood flow to bone. Therefore, it is possible that anemia reduces nutrient supply and predisposes bone to infarction, and that the distal areas of the long bones are susceptible because of decreased vascularity from small-sized vessels in subendochondral bone in these areas.
In this study, which was restricted to patients with GD1, we found proportionally fewer N370S homozygous patients among patients suffering from AVN compared to patients who did not experience this complication. However, N370S homozygous patients constituted 23.8% of all patients with AVN, confirming previous reports that, even in this apparently milder genotype, there is significant risk of AVN.3, 31 Consistent with our findings that the burden of visceral disease was not a predictor of AVN, Taddei and colleagues31 found an unexpectedly high incidence of AVN among older N370S homozygous patients, even though they exhibited minimal, or even no, visceral or hematologic disease. We found that, compared to N370S homozygous patients, AVN was more prevalent among heteroallelic genotypes (eg, N370S/L444P, N370S/84GG). Therefore, the risk of AVN occurs across all genotypes, irrespective of severity of visceral or skeletal disease. This is important in management of patients since the decision to start enzyme therapy should not be based solely on extent of visceral or hematological disease, but should take into consideration the risk of AVN.
This study was possible only because of the existence of a large longitudinal international disease registry, the ICGG Gaucher Registry, which provided a patient population sufficiently large for stratification and analysis by type of bone disease. The matched case-control analysis decreases bias in rare disease registry studies when data from multiple groups are compared.25 Yet, there are limitations associated with the use of ICGG Gaucher Registry data that are common to most observational (nonrandomized) research study data. All Registry data are retrospective and unaudited. Patients followed in the Registry are not randomized to treatment with imiglucerase. Other potential confounders of the impact of treatment with imiglucerase on bone mineral density, such as activity level, genetic predisposition, nutrition, vitamin D status, alcohol use, and smoking, were not considered in this analysis because these data are not recorded in the Registry. Additionally, geographic variations in sun exposure and differences in densitometry equipment (eg, Lunar, Hologic, or Norland), the norms from which the Z-scores were derived, were not taken into consideration for the same reasons.
In summary, low lumbar spine bone mineral density is a predictor of increased risk of fractures in GD1, and anemia emerged as the only risk factor for AVN. AVN and fractures are significant sources of long-term morbidity for patients with GD1. Imiglucerase therapy decreases the likelihood for the development of AVN10 and results in improved bone mineral density scores in many patients.12, 13, 20, 23 It will be of interest to determine whether the more potent and specific small-molecule substrate reduction therapy currently in clinical trials will result in additional therapeutic gains to further minimize the risk of these devastating complications of GD1.
AK, PKM, and NJW receive honoraria and expense reimbursement for serving on the Board of Advisors of the ICGG Gaucher Registry; travel reimbursements and/or honoraria and/or research support from Genzyme, a Sanofi company. NJW also receives honoraria/royalties from Shire Pharmaceuticals and Actelion; grant and contract support from Shire Pharmaceuticals and is a spokesperson or in the Speakers Bureau of Genzyme, a Sanofi company, Actelion and Shire Pharmaceuticals. AK and NJW do not hold any financial interest in any pharmaceutical company. TH receives travel reimbursement and/or honoraria for speaking engagements from Genzyme, a Sanofi company, and Shire Pharmaceuticals. JST is an employee of Genzyme, a Sanofi company. AK, PKM, NJW, and TH did not receive funding for this study.
Robert Brown, a graphic artist employed by Genzyme, a Sanofi company, helped design the tables. J Alexander Cole, DSc, MPH, contributed to the epidemiologic design of the analysis, in addition to drafting and editing the manuscript; he is an employee of Genzyme. Andrea Gwosdow, PhD, was responsible for writing, editing, and managing the outline and manuscript, and interpretation of data. This included managing author reviews and synthesizing the comments of each individual author into each draft of the manuscript. Andrea Gwosdow is a medical writer contracted by Genzyme. We thank the patients with type 1 (non-neuronopathic) Gaucher disease and their physicians and health care personnel who submitted data to the Gaucher Registry; the Gaucher Registry support team at Genzyme; members of the International Collaborative Gaucher Registry Board; and Radhika Tripuraneni, MD, Sarah Kulke, MD, and Gerald Cox, MD, PhD, for reviewing the manuscript. Logistical support for this work was provided by Genzyme. The database for the ICGG Gaucher Registry is supported by Genzyme.
Authors' roles: Study design: AK, TH, PKM, and NJW. Study conduct: AK, JST, and NJW. Data analysis: JST. Data interpretation: AK, TH, PKM, NJW, and JST. Drafting manuscript: JST, PKM, and NJW. Revising manuscript content: AK, TH, JST, PKM, and NJW. Approving final version of manuscript: AK, TH, JST, PKM, and NJW.