Fracture Risk Assessment (FRAX) without BMD and Risk of Major Osteoporotic Fractures in Adults with Type 1 Diabetes

Abstract

Objective

To evaluate the association between Fracture Risk Assessment Tool (FRAX) without bone mineral density (BMD) and risk for major osteoporotic fractures (MOF) in type 1 diabetes.

Methods

Subjects with type 1 diabetes and without diabetes from the ‘Coronary Artery Calcification in Type 1 Diabetes’ study were included. Risk for MOF was calculated using FRAX-based clinical risk factors and without BMD at visit 3 (2006–2008). Incident fractures were defined as fractures that occurred between visit 3 and visit 4 (2013–2017). Survival models were used to study the predictability of new MOF by diabetes status.

Results

346 type 1 diabetes (mean age 43.3 ± 9, BMI 26.4 ± 5, diabetes duration 29.4 ± 8.6 years, A1c 7.8 ± 1.1) and 411 controls (mean age 46.9 ± 9 years, BMI 26.3 ± 5 kg/m², A1c 5.5 ± 0.4) were analyzed in this study. In unadjusted survival analysis, the FRAX score without BMD was significantly associated with MOF (HR 1.08, 95% CI: 1.04–1.13, p<0.0001), and remained significantly associated after adjustment for age and sex (HR 1.09, 95% CI: 1.04–1.15, p=0.0007) and type 1 diabetes (HR 1.08, 95% CI: 1.04–1.12, p=0.0002). In the fully adjusted model (adjusted for age, sex and type 1 diabetes), the FRAX score without BMD was the only variable significantly associated with risk of MOF (HR 1.08, 95% CI: 1.02–1.14, p=0.006).

Conclusion

Clinical use of FRAX without BMD is useful tool in identifying adults with type 1 diabetes at higher risk for MOF risk and may help clinicians to guide therapeutic decision-making in this high fracture risk population.

1. Introduction

Osteoporotic fractures are increasingly recognized as an important complication of type 1 diabetes. Risk for major osteoporotic fractures is three-to four-fold higher, even in young adults with type 1 diabetes, compared to adults without diabetes.^{(1, 2)} There is a need for early screening and intervention in order to decrease the fracture morbidity and mortality in this population.

Areal bone mineral density (BMD) by dual x-ray absorptiometry (DXA) is one of the most widely used tools for estimation of fracture risk in the general population. A recent meta-analysis reported moderate reduction in femoral neck BMD and minimal or no change in BMD at the lumbar spine in adults with type 1 diabetes compared to non-diabetic adults.⁽³⁾ In a study by Vestergaard, the observed fracture risk among patients with type 1 diabetes was three times higher than calculated fracture risk based on DXA-measured BMD.⁽⁴⁾ This suggests that BMD by DXA underestimates the fracture risk in type 1 diabetes population. Trabecular bone score, a 2-dimentional gray scale technology to evaluate spine structural quality, has been shown to improve fracture risk prediction in type 2 diabetes population but has not been studied well in populations with type 1 diabetes.^{(5, 6)}

Besides BMD, several clinical factors such as age, sex, use of corticosteroids and certain chronic diseases are known to affect bone health and increase fracture risk. Therefore, the fracture risk assessment tool (FRAX) was developed to calculate the 10-year probability of major osteoporotic and hip fracture taking these clinical risk factors into account in addition to BMD. FRAX is a widely used clinical tool for fracture risk prediction and therapeutic decision making in many countries and is incorporated in many clinical guidelines.^(7–10) Studies in patients with type 2 diabetes have shown that FRAX underreports 10-year fracture probability for major osteoporotic fractures (MOF) and hip fractures^{(11, 12)}, probably due to increased BMD in patients with type 2 diabetes. Since type 2 diabetes is not a direct input in FRAX algorithm, a number of alternative strategies such as adjusting for diabetes related clinical factors, using low T-score for intervention and using rheumatoid arthritis have been suggested to improve FRAX performance.^{(6, 13, 14)}

Type 1 diabetes (formerly known as insulin dependent diabetes) is considered as one of the causes of secondary osteoporosis in the FRAX algorithm.⁽¹⁵⁾ However, type 1 diabetes is not a primary entry variable and it is given the same weight as other causes of secondary osteoporosis.⁽¹⁶⁾ Therefore, fracture probability increases when BMD is not included in the FRAX calculations.⁽¹⁶⁾ There are no studies on assessing FRAX-based fracture risk probability in adults with type 1 diabetes. We hypothesized that FRAX without BMD may provide accurate fracture risk prediction in adults with type 1 diabetes and hence, we evaluated the association of FRAX without BMD with fracture risk in a prospective study of patients with type 1 diabetes and controls without diabetes who were followed for more than 18 years at the Barbara Davis Center for Diabetes.

2. MATERIALS AND METHODS

2.1. Study Population

The Coronary Artery Calcification in Type 1 Diabetes (CACTI) study enrolled 1,416 participants, including both men and women with type 1 diabetes and without diabetes, between the ages of 20 and 55 at the baseline.⁽¹⁷⁾ Participants completed up to four visits over the course of 18 years, between 2000 and 2018. Type 1 diabetes was defined based on initiation of insulin therapy within a year of diabetes diagnosis and on insulin therapy at visit 3 (2006–2008) or diagnosed before the age of 30 or a clinical course consistent with type 1 diabetes.⁽¹⁷⁾ Nondiabetic control subjects were generally spouses, friends, and neighbors of the cases and had not been diagnosed with diabetes (including gestational) prior to enrollment and remained free of diabetes according to the American Diabetes Association criteria at visit 3 (2006–2008). At visit 4 (2013–2017), participants (n=805; 363 T1D, 442 controls) who responded to a fracture questionnaire were included in this study. The CACTI study is approved by Colorado Multiple Institutional Review Board and each subject signed informed written consent.

2.2. Fracture risk prediction tool

FRAX clinical factors such as age, sex, height, weight, BMI, history of previous fractures, parent hip fractures, smoking, use of corticosteroids, alcohol consumption and history of rheumatoid arthritis were used from the CACTI visit 3 (2006–2008). We entered type 1 diabetes as secondary osteoporosis in the FRAX model [15]. Height (in cm) was assessed with a wall mounted stadiometer and weight (in kg) was assessed without shoes using a digital scale. An average of two readings was taken as the final measurement. Weights of participants >125 kg was approximated to the upper cut off for weight in the FRAX (125kg) while calculating the scores. We defined previous fracture as any fracture that occurred before visit 3. Information on parental hip fracture was obtained from CACTI visit 3 questionnaires. If participants did not provide information on parental hip fracture, history of osteoporosis in either of the parents or both was used as a positive proxy for parental hip fracture. Current smoking history was obtained through self-report at visit 3. Participants with no smoking data were assumed as not current smokers. History of three or more units of daily alcohol consumption was obtained by self-report. As we did not have enough data on corticosteroid dose and duration of its use, we assumed it to be “none” for all participants uniformly in the study. The FRAX calculator did not have a provision for Native-Americans, hence, their scores were calculated using the tool for Caucasians. The FRAX scores for participants with no race information but who indicated non-Hispanic ethnicities were calculated using race information as Caucasians. The FRAX scores for participants below the age of 40 were calculated using their reported age, though the scores obtained are predictions at age 40.

2.3. FRAX Probabilities calculation

The probabilities for the major osteoporotic fractures and hip fractures were calculated using the FRAX tool (Web Version 4.1) for United States of America developed by the University of Sheffield (https://www.sheffield.ac.uk/FRAX), and included the clinical risk factors along with the variations as mentioned above.

2.4. Fracture Ascertainment

Incident fractures were identified as those that occurred between the participants’ age at visit 3 and visit 4, based on self-reported age at fracture. MOF included hip, forearm, humerus and spine. For individuals with more than one major osteoporotic fracture, analysis was based upon the time to first qualifying fracture.

2.5. Statistical analysis

Statistical analysis was performed with SAS 9.4 (SAS Institute Inc., Cary, NC). Descriptive statistics for demographic and baseline characteristics were compared by diabetes status and presented as mean ± standard deviation (SD) for continuous variables or number (%) for categorical variables. We considered diabetes as a binary variable (present vs absent). The FRAX scores without BMD were reported as mean ± SD. The categorical characteristics were compared using Chi-square tests. For small sample size (new hip fracture), Fisher’s Exact test was used.

Survival models were used to study the predictability of FRAX without BMD and type 1 diabetes on new MOF. The first model was unadjusted and then adjusted by age, and sex. In fully adjusted model, effect of FRAX without BMD on MOF was adjusted by age, sex and type 1 diabetes and effect of type 1 diabetes on MOF was adjusted by age, sex and FRAX without BMD. A sensitivity analysis was carried out by removing subjects younger than 40 years of age using similar unadjusted and adjusted survival models. All results are presented as hazard ratios (HR) with 95% confidence interval (CI).

3. Results

Out of the 805 participants who completed a fracture questionnaire at visit 4, participants with missing data required for the calculation of FRAX score (n=36), and controls who had developed T2D (n=12) at visit 3 were excluded from the study. Therefore, 757 participants (346 type 1 diabetes and 411 controls) were included in the analysis (Supplementary Figure 1). Of 346 participants with T1D, 143 were <40 years of age and of 411 controls, 91 were <40 years of age. Characteristics of the study population at visit 3 are shown in Table 1. There was no difference observed in the BMI between the two groups and as expected, HbA1c was higher in the type 1 diabetes group compared to the control group. The prevalence of rheumatoid arthritis (5.8% vs 1.2%, p=0.0003) and history of previous fractures (64.4% vs 56%, p=0.02) were significantly higher in the type 1 diabetes group compared to control group. Of 346 participants with type 1 diabetes, 162 had at least one self-reported diabetes microvascular complications such as retinopathy, neuropathy or nephropathy.

Table 1:

Baseline Characteristics of participants and FRAX-related clinical risk factors by diabetes status.

CHARACTERISTICS	Type 1 Diabetes Mellitus (n=346)	Non-Diabetic (n=411)	P-value
Age (Years)	43.3 ± 9	46.9 ± 9	<0.001
Diabetes Duration (Years)	29.4 ± 8.6	NA
BMI (kg/m2)	26.4 ± 5	26.3 ± 5	0.623
HbA1c (%)	7.8 ± 1.1	5.5 ± 0.4	<0.001
HbA1c (mmol/mol)	62	37	<0.001
Race*(n [%])			<0.001
Non-Hispanic White	334 (96.5)	351 (85.4)
Others
Hispanic	9 (2.6)	33 (8.0)
Black	3 (0.9)	19 (4.6)
Asian	0 (0.0)	8 (1.9)
≥ 3 units of alcohol/day (n [%])	12 (3.5)	5 (1.2)	0.036
Current Smoking (n [%])	21 (6.3)	18 (4.5)	0.284
Previous Fracture (n [%])	223 (64.4)	230 (56.0)	0.017
History of Osteoporosis (n [%])	27 (7.8)	16 (3.9)	0.021
Rheumatoid Arthritis (n [%])	20 (5.8)	5 (1.2)	<0.001

Statistics are presented as Mean ± SD, number (percentage). BMI; body mass index, HbA1c; Glycosylated hemoglobin, NA; Not Applicable.

^*Participants of Spanish origin were included under Hispanic, and participants who were of race other than those mentioned and not of Spanish origin, were included under Non-Hispanic White.

During an average of 8.5 ± 0.7 years of follow up for controls and 8.7 ± 0.8 years of follow up for type 1 diabetes, incident MOF was reported in 45 subjects (27 type 1 diabetes and 18 control) and incident hip fractures were reported in 6 subjects (5 type 1 diabetes and 1 control). The incidence of new MOF was significantly higher in type 1 diabetes when compared to control (7.8% vs 4.4%, p=0.048), while the incidence of new hip fracture was non-significantly greater in type 1 diabetes compared to controls (1.4% vs 0.2%, p=0.09).

The effect of type 1 diabetes and FRAX without BMD on incident MOF is shown in Figure 1. There was a significant association of type 1 diabetes with new incident MOF (HR 2.05, 95% CI: 1.11–3.76, p=0.02) when controlling for age and sex, but was not significant without adjustment (HR 1.82, 95% CI: 1.00–3.30, p=0.05). Moreover, the effect of type 1 diabetes on new incident MOF became statistically insignificant when adjusted with age, sex and FRAX without BMD (HR 1.57, 95% CI: 0.83–2.98, p=0.17) [Figure 1A]. We also examined the effect of 1% increase in FRAX score on MOF risk over 10 years. In unadjusted survival analysis, the FRAX score without BMD was significantly associated with MOF (HR 1.08, 95% CI: 1.04–1.13, p<0.0001), and remained significantly associated after adjustment for age and sex (HR 1.09, 95% CI: 1.04–1.15, p=0.0007) and type 1 diabetes (HR 1.08, 95% CI: 1.04–1.12, p=0.0002). In the model adjusted for age, sex, and type 1 diabetes, the FRAX score without BMD was the only variable significantly associated with risk of MOF (HR 1.08, 95% CI: 1.02–1.14, p=0.006) [Figure 1B]. Adjusting this model further for self-reported diabetes microvascular complications (retinopathy, neuropathy and nephropathy), FRAX without BMD was still significantly associated with MOF (HR 1.07; 95% CI: 1.02–1.13). Neither a composite complications variable nor the individual complications were associated with MOF.

Figure 1:

Effect of type 1 diabetes (A) and FRAX without BMD (B) on new major osteoporotic fractures

Statistics presented are Hazard Ratios and 95% Confidence Intervals (HR (95% CI)) and X-axis is labeled on logarithmic scale. MOF; Major osteoporotic fractures, T1D; Type 1 Diabetes; FRAX; fracture risk assessment tool, BMD; bone mineral density.

In a sensitivity analysis excluding subjects younger than 40 years of age, FRAX without BMD was significantly associated with new incidence MOF adjusted for type 1 diabetes (HR 1.08; 1.04–1.13, p=0.0004). There was no interaction for FRAX without BMD with type 1 diabetes and sex and type 1 diabetes with sex.

4. Discussion

To our knowledge, this is the first paper assessing the usefulness of FRAX without BMD for fracture risk prediction in patients with type 1 diabetes. Age and sex adjusted risk for a new MOF was two-fold higher even in this relatively young population with type 1 diabetes when compared to non-diabetics, which is in agreement with previous studies.^{(1, 2, 4)}. Association between type 1 diabetes and MOF became non-significant when adjusted for FRAX without BMD suggesting fracture risk in type 1 diabetes is captured well by FRAX without BMD. Moreover, FRAX without BMD was independently associated with new incident MOF in our cohort and it remained significant even after adjustment with age, sex, and type 1 diabetes and also after removing younger subjects aged 40 years and below.

In our study, FRAX without BMD identified fracture risk in adults with type 1 diabetes. This is in contrast with studies in type 2 diabetes populations where FRAX underestimated fracture risk.^{(11, 12)} In the FRAX algorithm, graded risk for secondary osteoporosis is based on graded risk for rheumatoid arthritis.⁽¹⁸⁾ We imputed type 1 diabetes as a cause of secondary osteoporosis in FRAX algorithm and therefore, the unadjusted FRAX without BMD was higher in type 1 diabetes when compared to non-diabetics and resulted in better fracture risk prediction in our cohort. However, BMD is a primary variable in the FRAX algorithm and therefore, entering BMD for FRAX would likely have underestimated fracture risk. As detailed in a review by Hough et al. ⁽¹⁶⁾, FRAX without BMD would estimate a fracture risk almost twice as high compared to FRAX with BMD in a 52-year-old postmenopausal woman without any other FRAX clinical risk factors.

The prevalence of rheumatoid arthritis was significantly higher in type 1 diabetes in our cohort. Studies have reported higher prevalence of rheumatoid arthirtis in people with T1D and it is likely due to a combination of genetic susceptibility and interactions between environmental risk factors and genes ⁽¹⁹⁾. In FRAX, the weight for rheumatoid arthritis and secondary osteoporosis are the same and the selection of rheumatoid arthritis while calculating the FRAX score does not take the secondary osteoporosis into account if selected simultaneously. Therefore, the higher prevalence of rheumatoid arthritis in our type 1 diabetes cohort would not have affected the findings of our study.

Some but not all studies suggest an association between diabetes microvascular complications and fracture risk ^(20,21). 50% of our T1D cohort had at least one diabetes microvascular complication. FRAX without BMD was a significant predictor of MOF despite adjusting for diabetes complications, and diabetes complications did not predict fracture risk in our population, suggesting that the increased fracture risk is not explained by complications.

Recent international consensus on diagnosis and management of bone fragility in diabetes suggested using the FRAX algorithm adjusted for rheumatoid arthritis to improve fracture risk prediction and make treatment decisions.⁽²²⁾ The consensus of use of adjusted FRAX was based on evidence from T2D populations. Our study in a type 1 diabetes population provides evidences for the utility of FRAX either adjusting for rheumatoid arthritis or using secondary osteoporosis without the use of BMD for fracture risk prediction.

This is the first study to assess FRAX without BMD in fracture risk prediction from an ongoing large cohort of patients with type 1 diabetes and controls of a similar age. Moreover, the CACTI study participants’ characteristics such as diabetes duration and glycemic control (HbA1c) was very similar to the US national T1D registry.⁽²³⁾ However, our study is not without limitations. Fractures were self-reported and not verified or adjudicated and self-reported fracture histories are subject to recall bias and may be subject to underreporting. Studies have reported good agreement between self-reported fractures and medical records of fractures ⁽²⁴⁾ and therefore, the findings of our study are likely to be valid with self-reported fractures. We used proxies for parental hip fracture because information on parental fractures was limited. However, number of proxies for parental hip fracture on both groups (T1D and controls) were similar and therefore, it may not have affected our findings. In the estimation of 10-year fracture probability, baseline covariates rather than time varying covariates were used as this was felt to be more relevant for the clinical assessment of fracture risk. Due to the younger age of our cohort and relatively shorter duration of follow-up, the incident hip fracture numbers were very small and therefore, we could not evaluate the utility of FRAX without BMD in hip fracture risk prediction.

In conclusion, findings of our study suggest using FRAX without BMD and using type 1 diabetes as a secondary osteoporosis for MOF risk estimation in adults with T1D may provide clinicians insight on making therapeutic decisions. Future prospective studies are needed to confirm our findings.

Highlights.

• Fracture risk is higher in people with T1D and this high fracture risk cannot be explained by DXA-measured BMD alone

• The fracture risk assessment tool (FRAX) underestimates fracture prediction in type 2 diabetes population but its utility has not been studied in T1D

• T1D was a significant predictor of new major osteoporotic fracture (MOF) adjusting for age and sex

• FRAX without BMD identifies adults with T1D at higher risk for MOF risk