Evaluation of Bone Mineral Density and Bone Biomarkers in Patients With Type 2 Diabetes Treated With Canagliflozin

Abstract

Context:

Canagliflozin is a sodium glucose cotransporter 2 inhibitor developed to treat type 2 diabetes mellitus (T2DM).

Objective:

Our objective is to describe the effects of canagliflozin on bone mineral density (BMD) and bone biomarkers in patients with T2DM.

Design:

This was a randomized study, consisting of a 26-week, double-blind, placebo-controlled period and a 78-week, double-blind, placebo-controlled extension.

Setting:

This study was undertaken in 90 centers in 17 countries.

Patients:

Patients were aged 55–80 years (N = 716) and whose T2DM was inadequately controlled on a stable antihyperglycemic regimen.

Interventions:

Canagliflozin 100 or 300 mg or placebo were administered once daily.

Outcome and Measures:

BMD was assessed using dual-energy x-ray absorptiometry at weeks 26, 52, and 104. Bone strength was assessed using quantitative computed tomography and finite element analysis at week 52. Serum collagen type 1 β-carboxy-telopeptide, osteocalcin, and estradiol were assessed at weeks 26 and 52.

Results:

Canagliflozin doses of 100 and 300 mg were associated with a decrease in total hip BMD over 104 weeks, (placebo-subtracted changes: −0.9% and −1.2%, respectively), but not at other sites measured (femoral neck, lumbar spine, or distal forearm). No meaningful changes in bone strength were observed. At week 52, canagliflozin was associated with an increase in collagen type 1 β-carboxy-telopeptide that was significantly correlated with a reduction in body weight, an increase in osteocalcin, and, in women, a decrease in estradiol.

Conclusions:

In older patients with T2DM, canagliflozin showed small but significant reductions in total hip BMD and increases in bone formation and resorption biomarkers, due at least in part to weight loss.

Canagliflozin is a sodium glucose cotransporter 2 (SGLT2) inhibitor that was developed for the treatment of adults with type 2 diabetes mellitus (T2DM) (1–9). Canagliflozin decreases plasma glucose in patients with T2DM by lowering the renal threshold for glucose and increasing urinary glucose excretion (10, 11). In phase 3 studies, canagliflozin was shown to improve glycemic control and reduce body weight and systolic blood pressure. It was generally well-tolerated in patients with T2DM as monotherapy or as an add-on agent to various antihyperglycemic agents (AHAs) (2–9).

Possible adverse effects of SGLT2 inhibitors on bone, including an increase in fractures and changes in bone biomarkers and bone mineral density (BMD), have been reported, yet additional studies are needed to determine the impact of SGLT2 inhibition on bone health (12). In toxicology studies, dose-related hyperostosis (increased trabecular bone content) was observed when canagliflozin treatment was initiated in younger rats (aged 6–8 weeks), but not when treatment was initiated in older rats (13). The increased trabecular bone content was normally mineralized. Canagliflozin is primarily an SGLT2 inhibitor, but based on findings from in vitro studies, it is also a weak inhibitor of sodium glucose cotransporter 1 (SGLT1) (14), the main transporter responsible for glucose absorption in the intestine (15). SGLT1 is expressed primarily in the kidney and intestine, whereas SGLT2 is expressed mainly in the kidney; neither transporter is expressed in bone (16, 17). During periods of drug absorption, intestinal concentrations of canagliflozin may become high enough to transiently inhibit SGLT1-mediated glucose transport (18), which contributes to the postprandial glucose reduction seen with canagliflozin treatment (19). Rats treated with canagliflozin showed symptoms of carbohydrate malabsorption as well as increased calcium solubility, enhanced calcium absorption, and hypercalciuria, likely leading to the observed skeletal hyperostosis (20, 21). As a result, absorptive hypercalciuria, reductions in hormones regulating calcium metabolism, and reductions in bone turnover markers, including parathyroid hormone (PTH) and 1,25-dihydroxyvitamin D, were observed in rats (13). Notably, hyperostosis was not observed in canagliflozin-treated mice or dogs. In humans, canagliflozin treatment was not associated with symptoms of carbohydrate malabsorption in clinical studies and there was no evidence of hypercalciuria (21).

In 12-week, phase 2 clinical studies, patients with T2DM and overweight/obese individuals without T2DM showed modest increases in the bone resorption marker collagen type 1 β-carboxy-telopeptide (β-CTX) with canagliflozin but no meaningful changes in other biomarkers (eg, serum and urine calcium and phosphate, PTH, 1,25-dihydroxyvitamin D, osteocalcin, propeptide amino terminal type 1 procollagen) (1, 22). Previous studies have demonstrated associations between weight loss and decreased levels of estradiol as well as increases in bone turnover markers and decreases in BMD (23–26). Given that reductions in body weight and estradiol were seen with canagliflozin in these phase 2 studies (1, 22), increases in β-CTX observed with canagliflozin may be due to weight loss, at least in part (27).

This manuscript describes the effects of canagliflozin on BMD and biochemical markers based on data from a 104-week, placebo-controlled, phase 3 clinical trial in patients with T2DM aged 55–80 years. The effects of canagliflozin on bone fracture risk are described in a separate report in this issue (28).

Materials and Methods

Patients and study design

An analysis of BMD and bone turnover markers was conducted as part of a double-blind, placebo-controlled, 104-week, phase 3 study (ClinicalTrials.gov Identifier: NCT01106651) evaluating the efficacy and safety of canagliflozin in patients aged 55–80 years with T2DM inadequately controlled on a range of background AHAs. Details of the study design have been described previously (5). Patients were randomized to canagliflozin 100 or 300 mg or placebo once daily during a 26-week treatment period followed by a 78-week extension period. Efficacy endpoints assessed at week 104 included change from baseline in hemoglobin A_1c (HbA_1c), fasting plasma glucose, systolic blood pressure, and percentage change from baseline in body weight and fasting plasma lipids. Safety and tolerability were also assessed. The overall efficacy and safety data have been reported separately (29).

Eligible patients were men and women aged 55–80 years with T2DM (women must have been ≥3 years postmenopausal), who had an HbA_1c of 7.0–10.0% and a body mass index (BMI) of 20–40 kg/m² at screening, and were either not on AHA therapy or were on a stable regimen of AHA(s) as monotherapy or combination therapy (including metformin, sulfonylurea, dipeptidyl peptidase-4 inhibitor, α-glucosidase inhibitor, glucagon-like peptide-1 agonist, or insulin [for ≥12 weeks before screening] or pioglitazone [for ≥6 months before screening] used in accordance with local prescribing information). Patients receiving medication to treat osteoporosis (eg, estrogen, selective estrogen receptor modulator [SERM] therapy, calcitonin) must have been on a stable treatment regimen for at least 6 months before screening. Patients using a bisphosphonate or those who were on PTH (1–34) (teriparatide) or denosumab treatment within 12 months before screening were excluded. Additional exclusion criteria included a T-score lower than −2.5 at any site (ie, lumbar spine, total hip, femoral neck, or distal 1/3 radius) in patients who were not on treatment with estrogen replacement, SERM, calcitonin, or other nonbisphosphonate therapy; severe vitamin D deficiency (serum 25-hydroxyvitamin D ≤10 ng/mL) at screening or within 12 months before screening; hypercalcemia at screening; conditions that might interfere with accurate measurement of BMD (eg, BMI >40 kg/m², severe scoliosis, degenerative spine changes, spinal fusion or metal implants, bilateral hip replacement, other surgery resulting in metal implants in both hips); a nonhealed fracture or any fracture within 12 months of screening; acquired or inherited bone disorders that may confound assessment of bone density or bone turnover (eg, Paget disease, osteomalacia, osteopetrosis, osteogenesis imperfecta); or elevated alkaline phosphatase activity (>1.5× upper limit of normal).

The study was conducted in accordance with ethical principles that comply with the Declaration of Helsinki, consistent with Good Clinical Practices and applicable regulatory requirements. The study protocol and amendments were approved by institutional review boards and independent ethics committees at participating institutions. All patients provided written informed consent before participation.

Randomization and treatments

Randomization and blinding details have been previously reported (5). Glycemic rescue therapy (up-titration of current AHA, except pioglitazone, or stepwise addition of AHAs, excluding peroxisome proliferator-activated receptor gamma [PPAR-γ] agents) was initiated in patients with fasting plasma glucose higher than 270 mg/dL (15.0 mmol/L) after day 1 to week 6, higher than 240 mg/dL (13.3 mmol/L) after week 6 to week 12, and higher than 200 mg/dL (11.1 mmol/L) after week 12 to week 26 or if HbA_1c was higher than 8.0% after week 26; investigators could initiate rescue therapy in patients with HbA_1c higher than 7.0% after week 52 if they determined that additional treatment was necessary.

Bone safety endpoints and analysis

BMD was assessed at the lumbar spine, hip (including femoral neck and total hip), and distal forearm (1/3 radius site) at baseline and weeks 26, 52, and 104 using dual-energy x-ray absorptiometry (DXA) with GE Lunar and Hologic scanners, with data acquired locally but analyzed by a central facility (BioClinica, Inc). BMD results obtained from GE Lunar machines were converted to Hologic values. Baseline Z-scores for BMD at the lumbar spine, distal forearm, femoral neck, and total hip were calculated. Volumetric BMD was assessed by quantitative computed tomography (QCT), and bone strength was assessed by QCT finite element analysis (FEA) at the spine and hip at baseline and week 52 in a subset of patients (n = 114) who were selected from study centers based on their expected rate of enrollment and the availability of technology and capabilities for the analysis. The strength to density ratio of the change in lumbar vertebra strength to volumetric BMD from baseline to week 52 was also assessed. All FEA image processing and analyses were performed at O.N. Diagnostics using VirtuOst software (version 1.1, O.N. Diagnostics) and blinded to treatment status. Construction of the finite element models has been described in detail elsewhere (30, 31). Vertebral measurements were taken from the L1 vertebra or the next available lumbar vertebra if the L1 vertebra was not suitable for analysis, as determined by the investigator or appropriately trained individual reviewing the scans. Hip measurements were taken from the left hip or the right hip if the left hip was not suitable for analysis. Estradiol (in women only), β-CTX, and osteocalcin were measured at baseline and weeks 26 and 52 using fasting serum samples stored at −70 °C. The analysis of bone safety included all randomized patients who took at least one dose of the study drug and was based on data regardless of the use of rescue medication. No hypothesis testing was specified for bone safety assessments. Summary statistics were calculated for the observed values and changes from baseline in bone parameters. Bone parameters were analyzed using analysis of covariance models with treatments, sex, baseline T-score category (<−1.5, ≥−1.5), and PPAR-γ use at baseline as fixed effects and baseline value as a covariate; least squares mean differences for each canagliflozin dose vs placebo and 95% confidence intervals (CIs) are reported. Comparisons of changes in bone parameters for which no hypothesis testing was prespecified and no adjustment for multiplicity was performed are denoted as statistically significant if the 95% CI for between-treatment differences excluded 0. Note that for all comparisons lacking prespecified hypothesis tests, this should be interpreted as nominal significance. Post hoc analyses were conducted to assess the change in total hip BMD adjusted for changes in body weight, and to evaluate the correlation between changes in β-CTX and changes in body weight.

Results

Patient disposition and baseline characteristics

A total of 716 patients were randomized into the study; 714 patients received at least one dose of the study drug. Of the 714 patients, 632 (88.5%) completed the 26-week core period and 520 (72.8%) completed 104 weeks of treatment. Baseline patient demographic and disease characteristics were generally similar across treatment groups, though there was a higher proportion of men in the placebo group compared with the canagliflozin groups (Table 1). Mean age was 63.6 years, BMI was 31.6 kg/m², and HbA_1c was 7.7%. Mean Z-scores for BMD at the lumbar spine, distal spine, femoral neck, and total hip were within the normal range across treatment groups. Most patients (88%) had T-scores of at least −1.5 at baseline; 12% of patients were on a PPAR-γ at baseline, and only 10 (1.4%) patients were on antiresorptive therapy at baseline (seven [1.0%] on estrogen, two [0.3%] on SERM, and one [0.1%] on bisphosphonate therapy).

Table 1.

Baseline Demographics and Disease Characteristics^*

Characteristic	PBO (n = 237)	CANA 100 mg (n = 241)	CANA 300 mg (n = 236)	Total (N = 714)
Sex, n (%)
Male	143 (60.3)	124 (51.5)	129 (54.7)	396 (55.5)
Female	94 (39.7)	117 (48.5)	107 (45.3)	318 (44.5)
Age, y	63.2 ± 6.2	64.3 ± 6.5	63.4 ± 6.0	63.6 ± 6.2
Race, n (%)†
White	185 (78.1)	194 (80.5)	173 (73.3)	552 (77.3)
Black or African American	20 (8.4)	18 (7.5)	19 (8.1)	57 (8.0)
Asian	21 (8.9)	15 (6.2)	25 (10.6)	61 (8.5)
Other‡	11 (4.6)	14 (5.8)	19 (8.1)	44 (6.2)
HbA1c, %	7.8 ± 0.8	7.8 ± 0.8	7.7 ± 0.8	7.7 ± 0.8
FPG, mg/dL (mmol/L)	157 ± 39 (8.7 ± 2.2)	160 ± 39 (8.9 ± 2.2)	153 ± 37 (8.5 ± 2.0)	157 ± 38 (8.7 ± 2.1)
Body weight, kg	91.1 ± 17.5	88.4 ± 15.6	88.8 ± 17.1	89.5 ± 16.8
BMI, kg/m2	31.8 ± 4.7	31.4 ± 4.4	31.5 ± 4.6	31.6 ± 4.6
Duration of T2DM, y	11.4 ± 7.3	12.3 ± 7.8	11.3 ± 7.2	11.7 ± 7.5
eGFR, mL/min/1.73 m2	76.1 ± 16.3	77.6 ± 17.0	78.7 ± 16.4	77.5 ± 16.6
Z-score
Lumbar spine§	1.0 ± 1.6	1.3 ± 1.4	1.3 ± 1.6	1.2 ± 1.5
Distal forearm‖	0.6 ± 1.2	0.7 ± 1.1	0.7 ± 1.2	0.7 ± 1.2
Femoral neck¶	0.5 ± 1.0	0.7 ± 1.0	0.7 ± 1.0	0.6 ± 1.0
Total hip¶	0.8 ± 1.1	1.0 ± 1.0	1.0 ± 1.1	0.9 ± 1.1

PBO, placebo; CANA, canagliflozin; FPG, fasting plasma glucose; BMI, body mass index; eGFR, estimated glomerular filtration rate; T2DM, type 2 diabetes mellitus.

^*Data are mean ± SD unless otherwise indicated.

^†Percentages may not total 100.0% due to rounding.

^‡Includes American Indian or Alaska Native, Native Hawaiian or other Pacific Islander, multiple, not reported, and other.

^§PBO, n = 227; CANA 100 mg, n = 228; CANA 300 mg, n = 227.

^‖PBO, n = 229; CANA 100 mg, n = 232; CANA 300 mg, n = 221.

^¶PBO, n = 230; CANA 100 mg, n = 230; CANA 300 mg, n = 227.

BMD

Total hip BMD decreased over 104 weeks in all treatment groups, with greater reductions observed with canagliflozin vs placebo at all time points (Figure 1A); at week 104, differences vs placebo were −0.9% and −1.2%, respectively. At other sites measured (lumbar spine, distal forearm, and femoral neck), no significant differences in changes from baseline to week 104 in BMD were observed between canagliflozin and placebo; 95% CIs included 0 for comparisons of both canagliflozin doses to placebo (Figure 1, B–D). The time course of BMD changes in the lumbar spine, distal radius, and femoral neck showed no increased separation in BMD changes from baseline over time between the canagliflozin and placebo groups.

Figure 1.

LS mean changes from baseline in BMD over 104 weeks for the (A) total hip, (B) lumbar spine, (C) distal forearm, and (D) femoral neck.^a,b BMD, bone mineral density; CANA, canagliflozin; CI, confidence interval; LS, least squares; PBO, placebo; SE, standard error. ^aGE Lunar BMD values were converted to Hologic units. ^bData are LS mean % change ± SE from baseline.

Canagliflozin doses of 100 and 300 mg were associated with reductions in body weight compared with placebo over 104 weeks (least squares mean percentage changes of −3.0% [−2.7 kg], −3.8% [−3.5 kg], and −0.6% [−0.6 kg], respectively) (29). To assess the potential contribution of weight loss to the observed reductions in total hip BMD in canagliflozin-treated patients, a post hoc analysis of changes from baseline in total hip BMD at week 104 was conducted that adjusted for changes in body weight. In this analysis, change in body weight was a statistically significant covariate (P = 1.2 × 10⁻⁸) and appeared to explain about 40% of the observed difference in total hip BMD between the pooled canagliflozin group and the placebo group (the weight change–adjusted difference in total BMD between pooled canagliflozin and placebo was −0.64% [P = .03] vs the unadjusted difference of −1.05%).

Volumetric BMD and bone strength

The QCT assessment of volumetric BMD at week 52 was consistent with the lack of changes seen with DXA at the spine and femoral neck. No significant changes in the ratio of lumbar vertebra strength to integral spine volumetric BMD, as measured by QCT and FEA, were seen with canagliflozin 100 and 300 mg vs placebo (0.2% and −0.2%, respectively; Supplemental Table 1).

Biomarkers

Changes in bone turnover markers associated with canagliflozin treatment over 52 weeks are summarized in Figure 2. Changes in serum estradiol were assessed in women (n = 301) who were not taking sex hormones (and excluding one outlier in the placebo group who had an increase from <120 pmol/L at baseline to 488 pmol/L at week 52 [1800% increase from baseline]). Relative to placebo, changes in estradiol with canagliflozin 100 and 300 mg were −4.4% and −13.7%, respectively, at week 26 and −14.2% and −21.0%, respectively, at week 52; the decrease in estradiol with canagliflozin 300 mg vs placebo was significant at both time points. Significant increases in the serum bone resorption marker, β-CTX, were seen with canagliflozin 100 and 300 mg relative to placebo at week 26 (+17.1% and +24.9%, respectively) and week 52 (+10.3% and +22.0%, respectively). Changes in the serum bone formation marker, osteocalcin, with canagliflozin 100 and 300 mg relative to placebo were +3.2% and +4.3%, respectively, at week 26 and +9.4% and +10.1%, respectively, at week 52; increases in serum osteocalcin with canagliflozin 100 and 300 mg vs placebo were significant at week 52.

Figure 2.

Changes from baseline in serum (A) estradiol,^a (B) β-CTX, and (C) osteocalcin at week 52.^b β-CTX, serum collagen type 1 β-carboxy-telopeptide; CANA, canagliflozin; CI, confidence interval; LS, least squares; PBO, placebo; SE, standard error. ^aIncludes only women and excludes one patient with an abnormal measurement. ^bData are LS mean % change ± SE from baseline.

Scatter plots demonstrating the correlation of changes in β-CTX with changes in body weight are shown in Figure 3. Changes in serum β-CTX were significantly correlated with reductions in body weight at week 26 (P < .001 for all treatment groups; placebo, r² = 0.073; canagliflozin 100 mg, r² = 0.092; canagliflozin 300 mg, r² = 0.104) and week 52 (placebo, P = .017, r² = 0.034; canagliflozin 100 mg, P = .007, r² = 0.036; canagliflozin 300 mg, P = .016, r² = 0.032), although as noted by the r² values, the changes in body weight explained only a small portion of the variability in serum β-CTX measurements.

Figure 3.

Scatter plot of change in serum β-CTX by change in body weight from baseline to (A) week 26 and (B) week 52. β-CTX, collagen type1 β-carboxy-telopeptide; CANA; canagliflozin; PBO, placebo.

Discussion

Canagliflozin treatment over 104 weeks was associated with a small (1.2%) but statistically significant reduction in total hip BMD compared with placebo; in the total hip, the SD is approximately 13%, so this change translates into a decrease of approximately 0.1 T-score units or 1% of peak bone mass. No significant changes in BMD were observed at other skeletal sites. At week 52, no significant changes in volumetric BMD or bone strength at the spine or femoral neck were observed, suggesting no impact on bone quality. Canagliflozin was associated with a significant increase in β-CTX at week 52 that was significantly correlated with weight loss; changes in bone resorption, which declined or remained steady after the initial assessment at week 26, were accompanied by significantly increased levels of osteocalcin, a measure of bone formation, over 52 weeks. The changes in total hip BMD at week 104 and in bone turnover and formation markers at week 52 that were seen with canagliflozin relative to placebo, in the context of no significant BMD changes in other skeletal sites (including the femoral neck) at week 104 or in volumetric BMD and bone strength at week 52, are of uncertain clinical significance.

In previous studies of weight loss, including diet- and exercise-induced weight loss, particularly in older patients, loss of weight has been associated with decreases in BMD (24, 26). This is, in part, due to the direct effect of reduced soft-tissue mass on bone through reduced mechanical loading (32). It has also been hypothesized that decreased fat tissue leads to reductions in aromatase activity, which lowers the production of estradiol and consequently increases bone turnover (33–35). Results from the current study are consistent with this hypothesis, based on previously reported changes in body fat with canagliflozin treatment (6, 36) and the changes in estradiol reported here. Furthermore, there was a statistically significant relationship between increases in serum β-CTX and weight loss. The small decrease in total hip BMD observed in this study is consistent with reductions in body weight leading to increased bone turnover. Approximately 40% of the reduction in total hip BMD with canagliflozin treatment was associated with weight loss. In previous studies that have shown a correlation of weight loss and decreased BMD, the extent of changes in both parameters was much greater than the changes reported here (26, 37–39). The reason for the change in BMD only at the total hip with canagliflozin treatment is not known. A recent study found that fat mass reductions in women correlated with changes in BMD only at the total hip and not at other sites (40). Thus, there might be site specificity to the effect of body weight reduction on BMD. Further studies are needed to define the mechanisms accounting for these differences.

A separate analysis reported in this issue found that canagliflozin was associated with an increased risk of fractures in the upper and lower extremities driven by a cohort of patients with a prior history or elevated risk of cardiovascular disease (28). The findings from that analysis suggest that an increased risk of fractures is due to factors not intrinsic to bone and that there is no clear relationship between fracture risk and overall bone health with canagliflozin treatment. Canagliflozin is not likely to have a direct effect on bone via SGLT2 inhibition because SGLT2 is not localized to bone or bone marrow (16, 17). An indirect effect of canagliflozin on bone through alterations in calcium homeostasis is also unlikely because canagliflozin treatment was also not associated with meaningful changes in serum calcium or phosphorus, PTH, and 25- or 1,25-dihydroxyvitamin D (1, 22).

Strengths of this study include the short- and long-term assessments of BMD using DXA at both central and peripheral skeletal locations in patients with T2DM. Further analysis integrating QCT and FEA technology allowed for an estimate of bone quality by examination of bone strength parameters that helped to better characterize the effects of canagliflozin on the structural properties of bone. One limitation of the study is that few patients with a baseline T-score below −2.5 were included, resulting in a slightly higher mean baseline than might be expected in the older T2DM patient population; of note, few patients were on background PPAR-γ or antiresorptive therapy. Additional limitations include the lack of assessment of gender-related differences in BMD because of the limited number of male and female patients, and that bone turnover markers and bone strength were not assessed at week 104. Moreover, these observations were based upon data captured during the study but not directed specifically to a hypothesis about canagliflozin and bone health. Lastly, approximately 70% of patients completed the 104-week treatment period; therefore, DXA BMD measurements were missing for the remaining ∼30%.

In conclusion, in patients aged 55–80 years with T2DM, canagliflozin treatment was associated with a decrease in total hip BMD and increases in bone resorption and formation biomarkers that are of uncertain clinical significance, particularly in the context of no significant BMD changes in other skeletal sites or in volumetric BMD and bone strength. The changes in total hip BMD as well as the increases in bone biomarkers observed with canagliflozin treatment may be, in part, related to weight loss.