You are here

Article Text

Article menu

Download PDFPDF

XML
Diabetes and endocrinology
Research
Associations with fracture in patients with diabetes: a nested case–control study
  1. Jakob Starup-Linde1,2,
  2. Søren Gregersen1,
  3. Peter Vestergaard2,3
  1. 1Department of Endocrinology and Internal Medicine (MEA), Aarhus University Hospital THG, Aarhus, Denmark
  2. 2Clinical Institute, Aalborg University, Aalborg, Denmark
  3. 3Department of Endocrinology, Aalborg University Hospital, Aalborg, Denmark
  1. Correspondence to Dr Jakob Starup-Linde; jakolind{at}rm.dk

Abstract

Objectives Diabetes mellitus is associated with an increased risk of fractures, which is not fully explained by bone mineral density and common risk factors. The aim of this study is to investigate the association of medication and biochemical markers on the risk of fracture in a diabetes population.

Design and setting A nested case–control study was conducted based on Danish diabetes patients from The Danish National Hospital Discharge Registry.

Participants The cases of the study were diabetes patients with a fracture (n=24 349), and controls were diabetes patients with no fracture (n=132 349). A total of 2627 diabetes patients were available for an analysis of patient characteristics, comorbidities, biochemical parameters and drug usage.

Results Age (OR=1.02, 95% CI 1.01 to 1.04), diabetes duration (OR=1.06, 95% CI 1.02 to 1.09), a diagnosis of previous fracture (OR=2.20, 95% CI 1.55 to 3.11), an alcohol-related diagnosis (OR=2.94, 95% CI 1.76 to 4.91), total cholesterol level (OR=2.50, 95% CI 1.20 to 5.21) and the usage of antiepileptics (OR=2.12, 95% CI 1.39 to 3.59) all increased the odds of fracture. Low-density lipoprotein cholesterol levels decreased the odds of fracture (OR =0.34, 95% CI 0.16 to 0.74), where the level of 3.04–5.96 mmol/L was optimal with regard to fracture risk.

Conclusions Low-density lipoprotein cholesterol may improve our understanding of fractures in diabetes patients, and it may be added to current fracture risk models in diabetes patients.

  • DIABETES & ENDOCRINOLOGY

This is an Open Access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/

https://doi.org/10.1136/bmjopen-2015-009686

Statistics from Altmetric.com

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

    Strengths and limitations of this study

    • The Danish National Hospital Discharge Registry covers the entire Danish population and has high validity.

    • Information on medication bought on prescription and clinically measured biochemical markers were available for a large group of patients with diabetes.

    • We were unable to assess whether medication was actually taken and at which intervals it was taken; however, we assume that non-compliance was only an issue in a small proportion, and that most antidiabetic agents and diabetes-associated therapies were taken on a regular basis.

    • This study was a retrospective case–control study with certain limitations, thus causality cannot be assessed.

    Introduction

    Diabetes mellitus (DM) is associated with an increased risk of fractures.1 ,2 The OR of hip fracture has been shown to be 6.9 (95% CI 3.3 to 14.8) in type 1 diabetes and 1.4 (95% CI 1.3 to 1.5) in type 2 diabetes.1 Owing to the high prevalence of type 2 diabetes in the general population, type 2 diabetes constitutes a large part of the total number of diabetes related fractures. Furthermore, bone mineral density (BMD) was found to be increased in type 2 diabetes, and decreased in type 1 diabetes. However, the decrease in BMD was not of a magnitude which can explain the increased fracture risk in type 1 diabetes.1 The Fracture Risk Assessment Tool (FRAX) score, a tool to determine fracture risk by BMD and common risk factors, for determining 10-year fracture risk was less valid in diabetes patients.3 Thus, common markers of bone frailty seem unable to detect and predict fractures in diabetes patients. Patients with diabetes may be more susceptible to falls due to hypoglycaemic events, orthostatic hypotension as an adverse effect to antihypertensive drugs, impaired vision and decreased sensation caused by retinopathy and neuropathy, foot ulcers and rapid fluctuations in plasma glucose.4 Observational studies report an increased risk of fracture when adjusted by hypoglycaemic events, previous falls and diabetes complications.5–7 Vestergaard et al8 did in a different cohort from the The Danish National Hospital Discharge Register report decreased fracture risk in metformin and sulfonylurea users. Other studies report neutral outcomes with the use of metformin and sulfonylurea.9–11 Glitazone use is reported to increase fracture risk in patients with diabetes.12 13 Previous observational studies have reported an increased risk of fracture with increasing HbA1c levels.14 ,15 The increased fracture risk in diabetes seems to be entangled in complications, medication use and biochemical markers.

    A previous observational study showed that low non-fasting high-density lipoprotein cholesterol (HDL) levels protected against fractures.16 However, low-density lipoprotein cholesterol (LDL) was not reported in this study, which may influence the interpretation of the results. Another observation found that total cholesterol, but not LDL or HDL,17 was a predictor of fracture risk, whereas, an additional study found no association of high total cholesterol and fractures.18

    The aim of this study is to investigate whether medication use and routine biochemical parameters are associated with fracture risk in diabetes patients.

    Methods

    The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement guideline for reports of case–control studies has been followed.19

    Design

    The study was conducted as a nested case–control study in a cohort of diabetes mellitus patients. The cases were diabetes mellitus patients with a subsequent fracture in the period 1 January 2008 to 31 December 2011. Controls were diabetes patients without a subsequent fracture in the same time period. A fracture before 2008 was considered as a previous fracture. Approval was obtained by the Danish Data Protection Agency.

    Diabetes assessment and fracture assessment

    We extracted data regarding all patients with DM from The Danish National Hospital Discharge Register in the time period 1 January 1977 to 31 December 2011 using International Classification of Diseases (ICD)-10 and ICD-8 codes (ICD-10 codes: E10-E14, ICD-8 codes: 249-250). The Danish National Hospital Discharge Register was founded in 1977 and covers all inpatient contacts from 1977 to 1994, and from 1995 also all outpatient visits to hospitals, outpatient clinics and emergency rooms. From the same register, we obtained incident fractures in the same time period using ICD-10 codes. (ICD-10 codes: S02.0–S02.9, S07.0–S07.9, S12.0–S12.9, S22.0–S22.9, S32.0–S32.8, S42.0–S42.9, S52.0–S52.9, S62.0–S62.9, S72.0–S72.9, S82.0–S82.9, S92.0–S92.9). In box 1 the fractures are grouped by site. The fractures were a sum of low-energy and high-energy fractures. It was possible to link a diagnosis of diabetes and diagnosis of fracture due to a unique personal identifier. The same unique identifier made it possible to link to the exposure variables below.

    Box 1

    Overview of included variables and respective codes from International Classification of Diseases (ICD), Anatomical Therapeutic Chemical Classification system (ATC), or Nomenclature for Properties and Units (NPU)

    Diabetes mellitus and Fractures (ICD)

    ▸ Type 1 diabetes mellitus: DE100, DE101, DE109, 24 900, 24 909, 24 907, 24 906

    ▸ Type 2 diabetes mellitus: DE110, DE111, DE119, 25 006, 25 007, 25 009

    Fractures

    ▸ Skull and facial bones: S02.0–S02.9, 800–804.99

    ▸ Neck: S12.0–S12.9

    ▸ Rib, sternum and thoracic spine: S22.0–S22.9, 807 809, 810

    ▸ Lumbar spine and pelvis: S32.0–S32.8, 808–808.09,808.11–808.19, 808.91–808.99, 809,

    ▸ Shoulder and upper arm: S42.0–S42.9, 811–812, 818–819

    ▸ Forearm: S52.0–S52.9, 813, 818–819

    ▸ Wrist and hand: S62.0–S62.9, 814–816.09, 816.19, 816.99–817, 818–819

    ▸ Femur: S72.0–S72.9, 820–820.12, 820.18–820.92, 820.98–821.22, 821.28–821.32, 821.38–821.92, 821.98–821.99, 827–829

    ▸ Lower leg and ankle: S82.0–S82.9, 822–824.03, 824.08–824.13, 824.18–824.93, 824.98–824.99, 827–829

    ▸ Foot: S92.0–S92.9, 825 826.01–826.19, 826.99–829.99

    ▸ Spine fractures (ICD 8): 805–806

    Comorbidities (ICD)

    ▸ AMI (DI21, DI22, DI23, 41 009, 41 099), diabetes-related nephropathy (DE102, DE112, 24 902, 25 002), diabetes-related neuropathy (DE104, DE114, 24 903, 25 003), diabetes-related retinopathy (DE103, DE113, 24 901, 25 001), diabetes-related peripheral artery disease (DE105, DE115, 24 904, 24 905, 25 004, 25 005), heart failure (DI50, DI110, DI130, DI132, 42 709 –42 711, 42 719, 42 899), alcohol (DF10, 303)

    Antidiabetic drugs, antihypertensive drugs, statins (ATC)

    ▸ Insulin (A10A), biguanides (A10BA), β-cell stimulating (A10BB), glitazones (A10BG), GLP-1 receptor agonists (A10BX) antihypertensive drugs (C09, C07A, C08, C02AB, C02AC, C02CA, C02DB, C03C, C03AA, C03D)

    statins (C10AA)

    Antiosteoporotic agents (ATC)

    ▸ Hormone replacement therapy (G03), strontium ranelate (M05BX03), teriparatide (H05AA), bisphosphonates (M05BA), denosumab (M05BX),

    Others (ATC)

    ▸ Antipsychotics (N05A), antiepileptics (N03A), antidepressants (N06A) glucocorticoids (H02AB)

    Biochemical markers (NPU)

    ▸ Alat (19 651), alkaline phosphatase (19 655), creatinine (18 016), HbA1c DCCT (03 835), total cholesterol (01 566), LDL cholesterol (01 568), HDL cholesterol (01 567), triglycerides (03 620), sodium (03 429), potassium (03 230), thyroid stimulating hormone (03 577), 25 hydroxy vitamin D (10 267)

    Exposure variables

    For an overview of acquired exposure variables and their related ICD-codes, Anatomical Therapeutic Chemical Classification system (ATC) codes, and Nomenclature for Properties and Units (NPU) codes (see box 1). NPU codes are an international classification system to be used in Clinical Laboratory Sciences. From The Danish National Hospital Discharge Register data regarding comorbidities were obtained. From the Central Region of Jutland, biochemical markers (2008–2011) and medication use from the prescription registry (2008–2011) by NPU-codes and ATC-codes were collected. Only data regarding redeemed drugs on prescription and not over-the-counter drugs were available. Age was defined as the age of the individual at 1 January 2008. Episodes of hypoglycaemia were collected. We did not limit the users by a time window, as hypoglycaemic events may be increased at the start of a new drug, and decreased bone strength would follow a longer period of use. The duration of DM (DM duration) was defined by the time from the date of DM diagnosis to the end of the prescription register, 31 December 2011. Furthermore, from the Central Region of Jutland BMI, and dual-energy X-ray absorptiometry (DXA), t-scores were collected, however, these were only available for a small subgroup and not applied in the adjusted analysis. Data on biochemical markers were collected from the hospitals in the Central Region of Jutland, thus, the number of patients with information differs by marker. LDL was calculated by the Friedwald formula. The Friedwald formula is less accurate for triglyceride levels of more than 2.3 mmol/L, and not applicable when levels are more than 4.5 mmol/L.20 We did not have data on smoking habits. Alcohol use was determined by the registration an alcohol-related diagnosis. A proxy variable for hypertension (yes/no) was created by drug usage; yes was defined as the use of diuretics, β-antagonists, calcium channel antagonists, ACE inhibitors/AT2 antagonists, antiadrenergic drugs with either peripheral or central effect, renin-antagonists, or hydralazine. A proxy variable for usage of drugs affecting bone was created by drug usage; yes was defined as the use of bisphosphonates, teriparatide, strontium ranelate, denosumab or hormone replacement therapy. Non-statin lipid-lowering drugs (fibrates, cholesterol absorption inhibitors, nicotinic acid and bile acid resins) were gathered in a composite exposure category.

    Data regarding pharmaceutical drug usage were collected in 29 929 individuals. Any drugs bought were registered with ATC code, dose sold and date of sale for the period 1 January 2008 to 31 December 2011. Because all sales were registered to the individual who redeemed the prescription, the capture and validity of data are high. Biochemical data and medication data were excluded if these were subsequent to entering case status, thus, biochemical data and medication use reflect the state before fracture. Means were calculated when more samples were available. The biochemical analyses were performed in an ISO standardised laboratory.

    The study group

    A total of 156 698 diabetes patients were collected; 29 929 of these had data on pharmaceutical use. All cases of nephropathy were excluded; 2627 diabetes patients had information on biochemical markers and a subset of these additional on thyroid stimulating hormones (TSH) and vitamin D status (n= 2412 and n= 1592; respectively). In total, 2627 diabetes patients were available for an analysis of patient characteristics, comorbidities, biochemical parameters and drug usage. Figure 1 depicts the selection of the patients.

    Figure 1
    Figure 1

    Flow chart of the selection of diabetes patients. Only patients with data on medication use and pharmaceutical use are included in the study.

    Statistical analysis

    The statistical analysis was conducted using the STATA V.8 statistics package. Pharmaceutical use was handled as (yes/no) and biochemical markers were handled as numerical values. An unadjusted case–control analysis, and an adjusted case–control analysis were performed using logistic regression. The adjusted analysis was performed to address potential confounding. The adjusted analysis included characteristics, pharmaceutical use and biochemical markers. All results are presented as OR. In the following, the expression risk will be used synonymously with OR as less than 10% of the study group are cases. Subgroup analysis was performed on individuals who provided data on biochemical markers and pharmaceutical drug use. Specific analysis was performed on type 1 and type 2 diabetes patients. A sensitivity analysis was conducted on patients with type 2 diabetes excluding those with an alcohol-related diagnosis. The adjusted subgroup analysis included an analysis grouped by LDL levels in eight quantiles. To ensure robustness of the results, we performed an analysis by LDL level in tertiles and quartiles for the entire model. Statin duration was defined as the time between first redeemed prescription and the time of event (for cases), or before the end of follow-up, 31 December 2011 (for controls). Separate analyses were conducted using statin duration as a variable and excluding statin use as a variable.

    Results

    A total number of 2627 diabetes mellitus patients were included in this study. A total of 175 (6.6%) of the diabetes patients had a subsequent fracture after the diabetes diagnosis. The 175 fractures were primarily located at the femur (19%), shoulder and upper arm (19%), forearm (17%), wrist and hand (12%), and knee and lower leg (12%), but also at the foot (5%), lumbar spine and pelvis (4%), rib, sternum and thoracic spine (4%), skull and facial bones (3%), and neck (2%). Table 1 gives an overview of evaluated characteristics, biochemical profile and pharmaceutical drug use of patients with and without fractures. The diabetes patients with fractures were significantly older, had a significantly longer diabetes duration, shorter duration of statin use, had a more frequent history of previous fracture and alcohol diagnosis, and cases were less frequent users of biguanides and glucagon-like peptide-1 (GLP-1) receptor agonists, and more frequent users of antidepressants and antiepileptics. Also, fewer of the patients with fracture used hydrophilic statins than patients without fracture, whereas, use of lipophilic statins did not differ between the groups. Patients with fractures were more susceptible to be users of loop diuretics compared to patients without fractures, whereas no difference was observed in thiazide use. Gender and diabetes complications were not significantly different between cases and controls.

    View this table:
    Table 1

    Baseline characteristics of the 2627 diabetes patients

    Table 2 presents the results of the unadjusted analysis. In the unadjusted analysis, increasing age (1.02; 95% CI 1.01 to 1.03), increasing diabetes duration (1.04; 95% CI 1.01 to 1.07), previous fracture (2.05; 95% CI 1.48 to 2.82), an alcohol related diagnosis (3.04; 95% CI 1.95 to 4.73) and neuropathy (1.92; 95% CI 1.11 to 3.32) were all associated with increased risk of fracture. The only biochemical markers associated with fracture risk were HDL (1.54; 95% CI 1.09 to 2.17) and potassium (0.55; 95% CI 0.33 to 0.91). Usage of the antidiabetics biguanides (0.68; 95% CI 0.49 to 0.93), and glucagon-like peptide-1 receptor agonists (0.31 95% CI 0.12 to 0.86), were associated with decreased fracture risk. Increasing duration of statin use was also associated with decreased fracture risk (0.83; 95% CI 0.76 to 0.91).

    View this table:
    Table 2

    Unadjusted and adjusted OR for fracture after diagnosis of diabetes n=2627

    Table 2 shows the results of the adjusted analysis. Increasing age (1.02; 95% CI 1.01 to 1.04), increasing diabetes duration (1.06; 95% CI 1.02 to 1.09), a previous fracture (2.20; 95% CI 1.55 to 3.11), and an alcohol-related diagnosis (2.94; 95% CI 1.76 to 4.91) preserves their significance. In this analysis, total cholesterol was associated with increased risk (2.50; 95% CI 1.20 to 5.21) and LDL (0.34; 95% CI 0.16 to 0.74) was associated with decreased risk of fracture significantly. Furthermore, potassium levels were associated with a decreased fracture risk (0.51; 95% CI 0.30 to 0.86). Antiepileptics were associated with increased fracture risk (2.12; 95% CI 1.33 to 3.59), whereas drugs affecting bone were associated with decreased fracture risk (0.50; 95% CI 0.27 to 0.95). Adjustment by statin duration instead of statin use did not alter significance of the results, with the exception of antihypertensive use, which was associated with increased fracture risk in this analysis (2.21; 95% CI 1.33 to 3.71). The adjusted analysis was performed specifically for type 1 and type 2 diabetes patients. LDL was still significantly associated with a decrease in fracture risk in the specific analysis for type 2 diabetes (selected results presented in table 2), and significance levels did not change. Too few type 1 diabetes patients were available for conducting a meaningful analysis (n=175). After exclusion of patients with an alcohol-related diagnosis, 1570 type 2 diabetes patients were available for analysis. In the fully adjusted model, LDL was not significantly associated with fracture (0.37; 95% CI 0.11 to 1.26). Likewise, total cholesterol was not significantly associated with fracture risk (2.41; 95% CI 0.74 to 7.80). Other variables did not change significance in this analysis.

    LDL and fracture risk

    In a stratified adjusted analysis specific to type 2 diabetes patients, the LDL level was divided by quantiles, and a significant association with increased risk of fracture was present for all quantiles when compared with the LDL level of 3.039–5.959 mmol/L. This reference group had a mean LDL level of 3.59 (0.52 SD). As statin use may influence the lipid levels, adjustment performed for statins was also included in the analysis (tables 2 and 3). Further adjustment by statin duration was performed, which did not change the results. Table 3 presents the relationship between LDL levels and fracture risk in type 2 diabetes patients.

    View this table:
    Table 3

    Multivariate adjusted OR for a fracture in type 2 diabetes patients by LDL levels, n=1663

    Table 4 presents the association between LDL levels divided in tertiles and quartiles, and fracture risk in the 2627 diabetes patients. The highest quartile of LDL was associated with a decreased fracture risk, when comparing with the reference level of 2.21–2.66 mmol/L LDL, whereas, the other quartiles did not differ from the reference. The lowest and highest tertile of LDL did not differ in comparison with the middle tertile.

    View this table:
    Table 4

    Multivariate adjusted OR for a fracture in diabetes patients by LDL levels in tertiles and quartiles, n=2627

    Discussion

    We found an association between risk of fracture and LDL-cholesterol for the whole cohort of diabetes patients as well as limited to type 2 diabetes. Nephropathy cases were excluded to eliminate bias from diabetic bone disease which range in a wide spectrum from adynamic bone disease to mixed renal osteodystrophy. Both low-energy and high-energy fractures were included in the analysis, as most fractures occur in non-osteoporotic individuals (T-score >−2.5).21 The usage of LDL levels as a predictor of fracture risk was limited to type 2 diabetes patients, as results for type 1 diabetes patients are of uncertain quality due to wide CIs. An LDL level of 3.04–5.95 mmol/L was associated with reduced fracture risk compared to lower LDL levels. The lowest LDL levels (0–1.5 mmol/L) were associated with an increased risk of fracture compared to higher levels. Thus, an LDL level of 3 mmol/L or more was associated with a decreased fracture risk. This was validated in a robustness test dividing LDL levels by quartiles, whereas the association was attenuated when dividing to tertiles. This may point at an optimal level of LDL in regard to fracture risk. This conflicts with current guidelines on LDL lowering in type 2 diabetes patients. The present American Diabetes Association (ADA) recommendations for LDL aim for levels at <1.8 mmol/L for participants with prior cardiovascular disease (CVD) and <2.5 for participants not having CVD.22 A review concluded that LDL lowering was beneficial for all but the lowest-risk type 2 diabetes patients in terms of CVD, though the optimal target value was not firmly established, and further research is needed.23 This study indicates that the CVD-preventive goal may increase the risk of fracture. This potential side effect should be assessed further in clinical trials determining the optimal target value of LDL in diabetes patients.

    Increasing diabetes duration and neuropathy were associated with increased fracture risk suggesting that diabetes severity was a determinant of fractures. In the population studied, neither retinopathy nor neuropathy was significantly associated with fracture in patients with diabetes. This is in line with a previous observational study, where the fracture risk in patients with diabetes was not associated with either neuropathy or retinopathy.7 Loop diuretics, but not thiazides, were more frequently used among patients with fracture compared to patients without fracture. This is consistent with previous findings.24 ,25 Furthermore, diabetes patients with fracture were less prone to be biguanide and GLP-1 receptor agonist users. The beneficial effect of biguanide and GLP-1 receptor agonist users may be explained by the insignificantly higher number of type 1 diabetes patients in the fracture group, as these are more prone to fractures. Also, caution should be taken when interpreting the GLP-1 receptor agonist as it came on the market in 2007 and only short-term use can be evaluated. Further investigations of the effects of severity, duration and treatment of diabetes are needed.

    Compared with other studies, this study may differ in the proportion of patients with type 1 DM, age and diabetes duration, incidence of retinopathy, metabolic control and use of antidiabetic drugs, which may have affected outcomes.6 ,7 ,26 A relatively large percentage of the diabetes patients had a previous alcohol-related diagnosis. The percentage was larger for those with a fracture (15.4%) than those without (5.7%). Adjustment was performed for the alcohol-related diagnosis, as it both may relate to diabetes and to fracture, and thus, confound the results. When excluding patients with an alcohol-related diagnosis, LDL was not significantly associated with fracture risk in type 2 diabetes patients, however, a strong trend was present. This suggests that the observed association between LDL and fracture may be caused by alcohol use and related liver disease as these are associated with fractures and low LDL levels.20 ,27 ,28 Furthermore, diabetes is associated with development of liver disease.29 However, we observed no association between biomarkers for liver and fracture risk. Further research is needed to determine whether LDL in regard of fracture is a marker of liver dysfunction.

    LDL may have a direct effect on bone. In vitro studies assessed the effect of LDL on bone cells; LDL-depletion in the culture medium was found to inhibit the formation of tartrate-resistant acid phosphatase multinucleated cells (osteoclast-like cells). The effect was abrogated with the addition of LDL to the suspension.30 The effect of LDL on osteoclasts was supported by an additional in vitro study, which found LDL to extend the survival of osteoclast-like cells, whereas HDL induces apoptosis.31 In addition, a study found the osteoclast-associated receptor, which is a costimulator of osteoclast differentiation, to be induced by oxidised LDL in endothelial cells; however, the effect was abolished by specific inhibition of the nuclear factor of activated T-cells.32

    An LDL receptor was found in osteoblasts.33 Human osteoblasts were found to undergo apoptosis when exposed to native and oxidised LDL. However, this study did not determine at which level LDL kills osteoblasts.34 A possible pathway of increased fracture risk with low LDL levels may also be found in the WNT-pathway. The low-density lipoprotein receptor-related proteins (LRP) 5, 6 and 8 were identified as positive regulators of the WNT-pathway and osteoblast differentiation.35 LDL binds to the LRP through apolipoprotein B and apolipoprotein E moieties,36 and may, therefore, stimulate the WNT-pathway. Cases of LRP 5 gene mutations with loss of function present with bone loss as well as cases of LRP 5 mutations present with abnormal glucose metabolism (diabetes and impaired glucose tolerance).37 In addition LRP 5 mutations may also affect lipid metabolism.37 This highlights the importance of a functioning LRP system. Consequently, too low concentrations of LDL may lead to impaired bone formation due to low WNT stimulation, which may explain the associations found in this study. We found no association between statin usage and fracture risk, however, increasing statin duration was associated with decreased fracture risk even when adjusting by LDL. The randomised JUPITER trial found no beneficial effect of statins and even a trend towards more fractures in the treated group.38 It highlights the potential dual effects of statins on bone: a beneficial effect through the mevalonate pathway39 and a negative effect through the LRP5 pathway.35

    The strengths of this study were that we evaluated a large group of diabetes patients and performed multiple adjustments by patient characteristics, comorbidities, biochemical markers and pharmaceutical drug use. The fully adjusted analysis revealed strong associations with potentially large clinical implications.

    However, this study was a retrospective case–control study with certain limitations, thus, causality cannot be assessed. This study may be influenced by selection bias, as biochemical markers only were available in a subpopulation. Sixty seven per cent of the diabetes patients were classified uniformly type 1 or type 2 diabetes, as some had an unspecified type and others were classified as both type 1 and type 2. The misclassification of diabetes diagnosis may limit our results. However, it may also reflect the difficulty in distinguishing type 1 and type 2 diabetes at the time of diagnosis. It is likely that 90% of the unspecified diabetes type is type 2 diabetes due to the population distribution. We were not able to distinguish between low-energy and high-energy trauma. This may affect our results, as high-energy trauma may be caused by unpredictable factors that do not relate to patient characteristics, medication use or biochemical markers. The inability to assess whether medication was actually taken, and at what intervals it was taken, is a limitation to the study. However, we assume that non-compliance was only an issue in a small proportion, and that most antidiabetic agents and diabetes-associated therapies were taken on a regular basis. The population was, in general, well regulated with HbA1c values within a narrow range. The tight glycaemic regulation supports good compliance with antidiabetic drugs. We also collected fall diagnoses, but few cases were recorded, so this did not change the results as only five of the 2627 patients had a prior diagnosis of tendency to falls, and none of these experienced a fracture. Previous studies show that the fracture risk is independent of previous falls.5 ,6 Subgroup analysis of LDL was performed by eight quantiles (n=207–209). Biochemical markers, pharmaceutical use, BMI and DXA hip t-score data were only available for a limited number of participants in the study. BMI was previously shown not to be significantly associated with LDL, and the lack of adjustment should thus not affect the results.40 Furthermore, important confounders like alcohol were evaluated through proxy variables. No data were available on smoking, and proxy variables performed poorly and were not used. Hormonal substitutions as growth hormone therapy and thyroid hormone therapy were not included in the analysis. However, TSH was evaluated in the analysis, and few participants were expected to receive growth hormone therapy. To our knowledge, we are the first to report an association between LDL and fracture risk in type 2 diabetes. The findings should be followed by clinical trials to decide whether LDL control should be performed in type 2 diabetes patients to prevent fractures and/or added to FRAX score in diabetes patients. In conclusion, the LDL level of 3.04–3.59 mmol/L was associated with a reduced risk of fracture compared to lower LDL levels.

    References

    1. ↵
      1. Vestergaard P
      . Discrepancies in bone mineral density and fracture risk in patients with type 1 and type 2 diabetes—a meta-analysis. Osteoporos Int 2007;18:427–44. doi:10.1007/s00198-006-0253-4
      OpenUrlCrossRefPubMedWeb of Science
    2. ↵
      1. Janghorbani M,
      2. Van Dam RM,
      3. Willett WC, et al
      . Systematic review of type 1 and type 2 diabetes mellitus and risk of fracture. Am J Epidemiol 2007;166:495–505. doi:10.1093/aje/kwm106
      OpenUrlAbstract/FREE Full Text
    3. ↵
      1. Giangregorio LM,
      2. Leslie WD,
      3. Lix LM, et al
      . FRAX underestimates fracture risk in patients with diabetes. J Bone Miner Res 2012;27:301–8. doi:10.1002/jbmr.556
      OpenUrlCrossRefPubMed
    4. ↵
      1. Mayne D,
      2. Stout NR,
      3. Aspray TJ
      . Diabetes, falls and fractures. Age Ageing 2010;39:522–5. doi:10.1093/ageing/afq081
      OpenUrlFREE Full Text
    5. ↵
      1. Bonds DE,
      2. Larson JC,
      3. Schwartz AV, et al
      . Risk of fracture in women with type 2 diabetes: the Women's Health Initiative Observational Study. J Clin Endocrinol Metab 2006;91:3404–10. doi:10.1210/jc.2006-0614
      OpenUrlCrossRefPubMedWeb of Science
    6. ↵
      1. Schwartz AV,
      2. Sellmeyer DE,
      3. Ensrud KE, et al
      . Older women with diabetes have an increased risk of fracture: a prospective study. J Clin Endocrinol Metab 2001;86:32–8. doi:10.1210/jcem.86.1.7139
      OpenUrlCrossRefPubMedWeb of Science
    7. ↵
      1. Vestergaard P,
      2. Rejnmark L,
      3. Mosekilde L
      . Diabetes and its complications and their relationship with risk of fractures in type 1 and 2 diabetes. Calcif Tissue Int 2009;84:45–55. doi:10.1007/s00223-008-9195-5
      OpenUrlCrossRefPubMedWeb of Science
    8. ↵
      1. Vestergaard P,
      2. Rejnmark L,
      3. Mosekilde L
      . Relative fracture risk in patients with diabetes mellitus, and the impact of insulin and oral antidiabetic medication on relative fracture risk. Diabetologia 2005;48:1292–9. doi:10.1007/s00125-005-1786-3
      OpenUrlCrossRefPubMedWeb of Science
    9. ↵
      1. Kanazawa I,
      2. Yamaguchi T,
      3. Yamamoto M, et al
      . Relationship between treatments with insulin and oral hypoglycemic agents versus the presence of vertebral fractures in type 2 diabetes mellitus. J Bone Miner Metab 2010;28:554–60. doi:10.1007/s00774-010-0160-9
      OpenUrlPubMed
    10. ↵
      1. Kahn SE,
      2. Zinman B,
      3. Lachin JM, et al
      . Rosiglitazone-associated fractures in type 2 diabetes: an Analysis from A Diabetes Outcome Progression Trial (ADOPT). Diabetes Care 2008;31:845–51. doi:10.2337/dc07-2270
      OpenUrlAbstract/FREE Full Text
    11. ↵
      1. Monami M,
      2. Cresci B,
      3. Colombini A, et al
      . Bone fractures and hypoglycemic treatment in type 2 diabetic patients: a case-control study. Diabetes Care 2008;31:199–203. doi:10.2337/dc07-1736
      OpenUrlAbstract/FREE Full Text
    12. ↵
      1. Bazelier MT,
      2. de Vries F,
      3. Vestergaard P, et al
      . Risk of fracture with thiazolidinediones: an individual patient data meta-analysis. Front Endocrinol (Lausanne) 2013;4:11. doi:10.3389/fendo.2013.00011
      OpenUrl
    13. ↵
      1. Loke YK,
      2. Singh S,
      3. Furberg CD
      . Long-term use of thiazolidinediones and fractures in type 2 diabetes: a meta-analysis. CMAJ 2009;180:32–9. doi:10.1503/cmaj.080486
      OpenUrlAbstract/FREE Full Text
    14. ↵
      1. Li CI,
      2. Liu CS,
      3. Lin WY, et al
      . Glycated hemoglobin level and risk of hip fracture in older people with type 2 diabetes: a competing risk analysis of Taiwan Diabetes Cohort Study. J Bone Miner Res 2015;30:1338–46. doi:10.1002/jbmr.2462
      OpenUrl
    15. ↵
      1. Neumann T,
      2. Lodes S,
      3. Kästner B, et al
      . High serum pentosidine but not esRAGE is associated with prevalent fractures in type 1 diabetes independent of bone mineral density and glycaemic control. Osteoporos Int 2014;25:1527–33. doi:10.1007/s00198-014-2631-7
      OpenUrlCrossRefPubMed
    16. ↵
      1. Ahmed LA,
      2. Schirmer H,
      3. Berntsen GK, et al
      . Features of the metabolic syndrome and the risk of non-vertebral fractures: the Tromso study. Osteoporos Int 2006;17:426–32. doi:10.1007/s00198-005-0003-z
      OpenUrlCrossRefPubMed
    17. ↵
      1. Trimpou P,
      2. Odén A,
      3. Simonsson T, et al
      . High serum total cholesterol is a long-term cause of osteoporotic fracture. Osteoporos Int 2011;22:1615–20. doi:10.1007/s00198-010-1367-2
      OpenUrlCrossRefPubMed
    18. ↵
      1. Dennison EM,
      2. Compston JE,
      3. Flahive J, et al
      . Effect of co-morbidities on fracture risk: Findings from the Global Longitudinal Study of Osteoporosis in Women (GLOW). Bone 2012;50:1288–93. doi:10.1016/j.bone.2012.02.639
      OpenUrlCrossRefPubMedWeb of Science
    19. ↵
      1. von Elm E,
      2. Altman DG,
      3. Egger M, et al
      . STROBE Initiative. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement: Guidelines for reporting observational studies. Int J Surg 2014;12:1495–9. doi:10.1016/j.ijsu.2014.07.013
      OpenUrlCrossRefPubMed
    20. ↵
      1. Sniderman AD,
      2. Blank D,
      3. Zakarian R, et al
      . Triglycerides and small dense LDL: the twin Achilles heels of the Friedewald formula. Clin Biochem 2003;36:499–504. doi:10.1016/S0009-9120(03)00117-6
      OpenUrlCrossRefPubMedWeb of Science
    21. ↵
      1. Siris ES,
      2. Miller PD,
      3. Barrett-Connor E, et al
      . Identification and fracture outcomes of undiagnosed low bone mineral density in postmenopausal women: results from The National Osteoporosis Risk Assessment. JAMA 2001;286:2815–22. doi:10.1001/jama.286.22.2815
      OpenUrlCrossRefPubMedWeb of Science
    22. ↵
      American Diabetes Association. Standards of medical care in diabetes—2008. Diabetes Care 2008;31(Suppl 1):S12–54. doi:10.2337/dc08-S012
      OpenUrlFREE Full Text
    23. ↵
      1. Rutter MK,
      2. Nesto RW
      . Blood pressure, lipids and glucose in type 2 diabetes: how low should we go? Re-discovering personalized care. Eur Heart J 2011;32:2247–55. doi:10.1093/eurheartj/ehr154
      OpenUrlAbstract/FREE Full Text
    24. ↵
      1. Kruse C,
      2. Eiken P,
      3. Vestergaard P
      . Continuous and long-term treatment is more important than dosage for the protective effect of thiazide use on bone metabolism and fracture risk. J Intern Med 2016;279:110–22. doi:10.1111/joim.12397
      OpenUrl
    25. ↵
      1. Xiao F,
      2. Qu X,
      3. Zhai Z, et al
      . Association between loop diuretic use and fracture risk. Osteoporos Int 2015;26:775–84. doi:10.1007/s00198-014-2979-8
      OpenUrl
    26. ↵
      1. Napoli N,
      2. Strotmeyer ES,
      3. Ensrud KE, et al
      . Fracture risk in diabetic elderly men: the MrOS study. Diabetologia 2014;57:2057–65. doi:10.1007/s00125-014-3289-6
      OpenUrlCrossRefPubMed
    27. ↵
      1. Nakchbandi IA
      . Osteoporosis and fractures in liver disease: relevance, pathogenesis and therapeutic implications. World J Gastroenterol 2014;20:9427–38. doi:10.3748/wjg.v20.i28.9427
      OpenUrl
    28. ↵
      1. Ghadir MR,
      2. Riahin AA,
      3. Havaspour A, et al
      . The relationship between lipid profile and severity of liver damage in cirrhotic patients. Hepat Mon 2010;10:285–8.
      OpenUrl
    29. ↵
      1. Tolman KG,
      2. Fonseca V,
      3. Dalpiaz A, et al
      . Spectrum of liver disease in type 2 diabetes and management of patients with diabetes and liver disease. Diabetes Care 2007;30:734–43. doi:10.2337/dc06-1539
      OpenUrlFREE Full Text
    30. ↵
      1. Sato T,
      2. Morita I,
      3. Murota S
      . Involvement of cholesterol in osteoclast-like cell formation via cellular fusion. Bone 1998;23:135–40. doi:10.1016/S8756-3282(98)00082-9
      OpenUrl
    31. ↵
      1. Luegmayr E,
      2. Glantschnig H,
      3. Wesolowski GA, et al
      . Osteoclast formation, survival and morphology are highly dependent on exogenous cholesterol/lipoproteins. Cell Death Differ 2004;11(Suppl 1):S108–18. doi:10.1038/sj.cdd.4401399
      OpenUrlCrossRefPubMed
    32. ↵
      1. Goettsch C,
      2. Rauner M,
      3. Sinningen K, et al
      . The osteoclast-associated receptor (OSCAR) is a novel receptor regulated by oxidized low-density lipoprotein in human endothelial cells. Endocrinology 2011;152:4915–26. doi:10.1210/en.2011-1282
      OpenUrlCrossRefPubMedWeb of Science
    33. ↵
      1. Dong Y,
      2. Lathrop W,
      3. Weaver D, et al
      . Molecular cloning and characterization of LR3, a novel LDL receptor family protein with mitogenic activity. Biochem Biophys Res Commun 1998;251:784–90. doi:10.1006/bbrc.1998.9545
      OpenUrlCrossRefPubMed
    34. ↵
      1. Klein BY,
      2. Rojansky N,
      3. Ben-Yehuda A, et al
      . Cell death in cultured human Saos2 osteoblasts exposed to low-density lipoprotein. J Cell Biochem 2003;90:42–58. doi:10.1002/jcb.10603
      OpenUrlCrossRefPubMedWeb of Science
    35. ↵
      1. Zhang J,
      2. Zhang X,
      3. Zhang L, et al
      . LRP8 mediates Wnt/beta-catenin signaling and controls osteoblast differentiation. J Bone Miner Res 2012;27:2065–74. doi:10.1002/jbmr.1661
      OpenUrlPubMed
    36. ↵
      1. Beisiegel U,
      2. Weber W,
      3. Ihrke G, et al
      . The LDL-receptor-related protein, LRP, is an apolipoprotein E-binding protein. Nature 1989;341:162–4. doi:10.1038/341162a0
      OpenUrlCrossRefPubMed
    37. ↵
      1. Saarinen A,
      2. Saukkonen T,
      3. Kivelä T, et al
      . Low density lipoprotein receptor-related protein 5 (LRP5) mutations and osteoporosis, impaired glucose metabolism and hypercholesterolaemia. Clin Endocrinol (Oxf) 2010;72:481–8. doi:10.1111/j.1365-2265.2009.03680.x
      OpenUrlCrossRefPubMed
    38. ↵
      1. Peña JM,
      2. Aspberg S,
      3. MacFadyen J, et al
      . Statin Therapy and Risk of Fracture: Results From the JUPITER Randomized Clinical Trial. JAMA Intern Med 2014;175:171–7. doi:10.1001/jamainternmed.2014.6388
      OpenUrl
    39. ↵
      1. Bauer DC
      . HMG CoA reductase inhibitors and the skeleton: a comprehensive review. Osteoporos Int 2003;14:273–82. doi:10.1007/s00198-002-1323-x
      OpenUrlCrossRefPubMedWeb of Science
    40. ↵
      1. Shamai L,
      2. Lurix E,
      3. Shen M, et al
      . Association of body mass index and lipid profiles: evaluation of a broad spectrum of body mass index patients including the morbidly obese. Obes Surg 2011;21:42–7. doi:10.1007/s11695-010-0170-7
      OpenUrlPubMed

    Footnotes

    • Contributors SG and PV planned the study. JS-L and PV researched data and conducted the statistics. JS-L, SG and PV wrote the manuscript. PV is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

    • Funding This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.

    • Competing interests None declared.

    • Ethics approval Danish Data Protection Agency.

    • Provenance and peer review Not commissioned; externally peer reviewed.

    • Data sharing statement No additional data are available.