Abstract
Context:
In trials, thiazolidinediones (TZDs) increase fracture risk in women, but the effects of discontinuation are unknown.
Objective:
The objective was to investigate the effects of TZD use and discontinuation on fractures in women and men.
Design:
This was a longitudinal observational cohort study using data from the Action to Control Cardiovascular Risk in Diabetes (ACCORD) trial bone ancillary study. Duration of TZD use and discontinuation during ACCORD, assessed every 2–4 months at clinic visits, were modeled as time-varying covariates in proportional hazards models for occurrence of first non-spine fracture.
Participants:
We studied a total of 6865 participants in ACCORD BONE.
Main Outcome Measures:
Main outcome measures were centrally adjudicated non-spine fracture.
Results:
Average age was 62.4 (SD, 6.6) years; average duration of diabetes was 11.1 (SD, 7.8) years. Rosiglitazone was used by 74% and pioglitazone by 13% of participants. During a mean follow-up of 4.8 (SD, 1.5) years, 262 men and 287 women experienced at least one non-spine fracture. The fracture rate was higher in women with 1–2 years of TZD use (hazard ratio [HR] = 2.32; 95% confidence interval [CI], 1.49, 3.62) or >2 years of TZD use (HR = 2.01; 95% CI, 1.35, 2.98), compared with no use. The fracture rate was reduced in women who had discontinued TZD use for 1–2 years (HR = 0.57; 95% CI, 0.35, 0.92) or > 2 years (HR = 0.42; 95% CI, 0.24, 0.74) compared with current users. TZD use and discontinuation were not associated with non-spine fractures in men.
Conclusions:
TZD use was associated with increased non-spine fractures in women, but not men, with type 2 diabetes. When women discontinued TZD use, the fracture effects were attenuated.
Fractures in older adults are associated with substantial morbidity and mortality (1) and are more common among those with diabetes (2). In this context, there has been particular concern about thiazolidinediones (TZDs), which have a negative effect on the skeleton. In post hoc analyses of randomized controlled trials of TZDs, the currently available TZDs, rosiglitazone and pioglitazone, increased fracture risk in women but not in men (3, 4). Some, but not all, observational studies have reported increased fracture risk with TZD use in men (5, 6).
There is some evidence regarding the effects of TZD discontinuation on bone mineral density (BMD). In a rodent model (7) and a clinical trial (8), the more rapid bone loss associated with rosiglitazone use was attenuated after use was discontinued. However, the effects of discontinuation on the key outcome of fracture risk are not known.
To investigate the effects of TZD use and discontinuation on fractures in women and men, we used data from an ancillary study of skeletal health (9) within the Action to Control Cardiovascular Risk in Diabetes (ACCORD) trial, ACCORD BONE (ClinicalTrials.gov Identifier: NCT00324350; https://clinicaltrials.gov/ct2/show/NCT00324350?term=nct00324350&rank=1).
Materials and Methods
ACCORD trial design and participants
The ACCORD trial has been described previously (10). Briefly, inclusion criteria for participants were: presence of type 2 diabetes; a hemoglobin A1c (A1C) of 7.5 to 11%; and age of 40–79 years with a history of cardiovascular disease (CVD), or 55–79 years with subclinical evidence of CVD or significant risk factors for CVD. Exclusion criteria included: frequent serious hypoglycemia; body mass index (BMI) of 45 kg/m2 or more; serum creatinine above 1.5 mg/dL (132.6 μmol/L); or other serious illness. The trial was conducted at 77 clinical sites in the United States and Canada, grouped into seven clinical center networks.
ACCORD was a double two-by-two factorial, parallel treatment trial. All eligible participants (n = 10 251) were randomly assigned to the intensive or standard glycemia therapy group. Participants were also assigned to either the blood pressure (BP) (n = 4733) or lipid (n = 5518) trial, based on eligibility. Randomization of 1174 participants occurred during the vanguard phase of the trial in 2001–2002. The remaining 9077 participants were randomized from January 1, 2003, to October 29, 2005. Randomization to intensive or standard glycemic control was stratified by clinical center network during the vanguard phase and by clinical site during the remainder of the trial. During the trial, ACCORD achieved a median A1C of 6.4 and 7.5% in the intensive and standard glycemia therapy groups, respectively. The intensive glycemia intervention protocol was stopped, and all participants were assigned to the standard glycemia protocol in February 2008 because of higher all-cause mortality in those assigned to the intensive treatment strategy (11). The trial ended in June 2009.
TZD use
At the baseline visit, participants were queried regarding their current use of diabetes medications including TZDs; dose and previous duration of use were not recorded. However, participants were required to have stable diabetes therapy for at least 3 months before enrollment in ACCORD. During the trial, rosiglitazone was included in the original treatment algorithm and was provided by the ACCORD formulary for the intensive and standard glycemia therapy groups. Pioglitazone could be used if available to the participant and was added to the ACCORD formulary in Fall 2007. Rosiglitazone was the more widely used TZD in ACCORD. Participants in the intensive glycemia control arm had visits every 2 months, whereas those in the standard control arm had visits every 4 months. At each visit, staff recorded the participant's use of diabetes medications, including TZDs, at visit entry. At visit exit, the currently prescribed medications were recorded. Compliance was assessed qualitatively at each visit, but a pill count was not performed.
BONE ancillary study design
Fracture ascertainment was conducted among participants in the BONE ancillary study. Five of the seven clinical center networks participated in ACCORD BONE, comprising 54 of 77 clinical sites. The BONE ancillary study was initiated during the main recruitment for the ACCORD trial. Beginning in January 2006, participants were asked at the next annual visit about the occurrence of any non-spine fractures since randomization. After the annual visit in 2006, participants were asked at each annual visit if they had experienced a fracture since their last annual visit. Of the 7287 randomized participants at the ACCORD BONE clinics, 6868 answered at least one question regarding fractures during ACCORD and were eligible for these analyses. Fractures of the lumbar and thoracic spine were excluded because of the additional burden of adjudicating such fractures. Reported fracture events were centrally adjudicated, based on radiology records, at the University of California, San Francisco with the adjudicators blinded to treatment assignment. A sample of reported events was independently adjudicated by a panel of experts at two timepoints during the trial for quality assurance. Only confirmed non-spine fractures were included in these analyses. As per protocol, pathological fractures, confirmed as occurring secondary to neoplasm, necrosis, or sepsis, and periprosthetic fractures were excluded (n = 7). The main outcome was all non-spine fractures. For these analyses, we also classified fractures by location as upper extremity (eg, hand, distal forearm, proximal humerus), lower extremity (eg, foot, ankle, leg, hip), or axial (rib, chest/sternum, face). Standing height was measured at baseline and annual follow-up visits on all ACCORD participants. Institutional review boards at all participating institutions approved the protocol for the ACCORD BONE ancillary study, and written informed consent was obtained from all participants.
Statistical methods
“TZD use” included cumulative use of rosiglitazone or pioglitazone during ACCORD, from randomization until the participant's last clinic visit before a fracture event occurred or the trial ended. Baseline use was considered as a separate variable and was not included in the calculation of TZD use. TZD use was time-varying and was categorized as no use, use for no more than 1 year, use for more than 1 year but no more than 2 years, and use for more than 2 years. The first cutpoint of 1 year was chosen based on evidence from the ADOPT (A Diabetes Outcome Progression Trial) trial that the fracture rate begins to increase after 1 year of TZD use (12). TZD discontinuation was also time-varying; was based on time since last reported use during ACCORD, up to the last clinic visit before a fracture event or the end of the trial, excluding baseline use; and included four categories: current user, 1 year or less, more than 1 year but less than 2 years, and more than 2 years since discontinuation (last use). Time-varying variables for rosiglitazone use and discontinuation were constructed similarly. Pioglitazone use was less common during ACCORD; time-varying variables were defined for “never,” “past,” and “current” pioglitazone use.
Cox proportional hazards models were used to assess the associations of TZD use and TZD discontinuation with the time until first nonpathological non-spine fracture. Mixed effect models were used to assess these associations with the outcome of change in height from baseline. Initial models were adjusted for age and glycemia intervention assignment. Multivariable models were constructed with these two variables and additional adjustment for the following baseline variables that have been associated with fracture and/or TZD use in previous studies: race/ethnicity, BMI, insulin use, TZD use, diabetes duration, A1C, history of congestive heart failure, smoking status, and assignments to BP or lipid trial and interventions. There is evidence from observational studies that insulin use is associated with fracture risk (13), and baseline insulin use was associated with incident fracture in ACCORD. Results from the ADOPT trial indicate that metformin and sulfonylurea use are not associated with fracture risk (12), and baseline use of these medications was not associated with incident fracture in ACCORD. Models assessing the effects of rosiglitazone use and discontinuation were additionally adjusted for pioglitazone use. Gender by TZD use interaction was examined. Net effects of TZD use and TZD discontinuation on fracture were calculated using linear contrasts. Analyses were performed at Wake Forest School of Medicine using SAS version 9.3 (SAS Institute).
Results
Of the 6868 ACCORD BONE participants with data on fracture occurrence, three did not have data on medication use during follow-up and were excluded from these analyses. Among the 6865 participants included in these analyses, 4494 (65%) were men (Table 1). Mean age was 62.4 (SD, 6.6) years; mean BMI was 32.6 (SD, 5.3) kg/m2; and mean duration of diabetes was 11.1 (SD, 7.8) years. At baseline, 20% of participants reported TZD use. During ACCORD, 76% of participants had at least one recorded period of TZD use. Rosiglitazone alone was used by 63% of participants, a combination of rosiglitazone and pioglitazone by 12%, and pioglitazone alone by 1%. Among those with any TZD use during ACCORD, the median duration of use was 3.2 (interquartile range, 1.8–4.3) years for men and 2.7 (1.2–3.9) years for women.
Table 1.
Baseline Characteristics of ACCORD BONE Participants
| Characteristic | Total | Men | Women |
|---|---|---|---|
| n | 6865 | 4494 | 2371 |
| Age, y | 62.4 (6.6) | 62.7 (6.7) | 62.0 (6.4) |
| Race/ethnicity | |||
| White | 4755 (69.3) | 3255 (72.4) | 1500 (63.3) |
| Latino | 174 (2.5) | 132 (2.9) | 42 (1.8) |
| Black | 1392 (20.3) | 768 (17.1) | 624 (26.3) |
| Asian | 125 (1.8) | 83 (1.9) | 42 (1.8) |
| Other | 419 (6.1) | 256 (5.7) | 163 (6.9) |
| Smoking | |||
| Never | 2574 (37.5) | 1309 (29.2) | 1265 (53.4) |
| Past | 3342 (48.8) | 2526 (56.3) | 816 (34.5) |
| Current | 940 (13.7) | 653 (14.6) | 287 (12.1) |
| BMI, kg/m2 | 32.5 (5.3) | 31.9 (4.9) | 33.8 (5.7) |
| Diabetes duration, y | 11.1 (7.8) | 11.0 (7.7) | 11.2 (7.8) |
| Baseline insulin use | 2569 (37.4) | 1587 (35.3) | 982 (41.4) |
| Baseline TZD use | 1344 (19.6) | 912 (20.3) | 432 (18.2) |
| Intensive glycemia intervention arm | 3408 (49.6) | 2224 (49.5) | 1184 (49.9) |
Data are expressed as number (percentage) or mean (SD).
During mean follow-up of 4.8 (SD, 1.5) years, 549 participants (262 or 5.8% men, and 287 or 12.1% women) experienced at least one confirmed non-spine fracture. The most common fractures were: in men, rib (20%), ankle (18%), toe (8%), leg (8%), and proximal humerus (7%); and in women, ankle (19%), proximal humerus (13%), foot (13%), distal forearm (10%), and leg (9%). There were 18 and 20 hip fractures in men and women, respectively. In men, there were 76 upper extremity, 134 lower extremity, and 72 axial fractures; in women, there were 106, 175, and 31, respectively. Because vertebral fractures were not identified in ACCORD, axial fractures included primarily rib fractures as well as smaller numbers of chest/sternum, tailbone, and facial fractures.
There was a statistically significant interaction between gender and duration of TZD use in multivariable-adjusted proportional hazard models for non-spine fracture (P < .001). Results were therefore analyzed separately for men and women. In models that were minimally adjusted for age and glycemia randomization assignment, the non-spine fracture rate was increased with TZD use (P < .01) and decreased with discontinuation (P = .03) in women, but not men (P = .24 and 0.51, respectively). Results were similar after adjustment for baseline TZD use and additional baseline variables: race/ethnicity, BMI, insulin use, diabetes duration, A1C, history of congestive heart failure, smoking status, and assignments to BP or lipid trial and interventions (Table 2). In men, the overall P value for an effect of TZD use on fracture was .17, indicating no statistically significant differences in fracture rate across the four categories of TZD use, despite the P value of .04 for TZD use of 0–1 year compared with no TZD use (14). The fracture rate was more than doubled in women with 1–2 years of TZD use (hazard ratio [HR] = 2.32; 95% confidence interval [CI], 1.49, 3.62) or > 2 years of TZD use (HR = 2.01; 95% CI, 1.35, 2.98), compared with no use. The fracture rate was reduced in women who discontinued TZD use compared with women who were current users. For women who had discontinued use for > 2 years, the HR was 0.42 (95% CI, 0.24, 0.74), compared with current users. The net effects of duration of TZD use and discontinuation on non-spine fracture rates in women were compared with “never use” as the reference group (Figure 1). The results indicate that the fracture rate after TZD discontinuation returns to the same rate as that of subjects who never used a TZD. For example, the HR for non-spine fracture was 0.97 (95% CI, 0.50, 1.88) comparing women with 1–2 years of TZD use who had discontinued for more than 2 years with women who never used a TZD.
Table 2.
Adjusteda HRs for Non-Spine Fracture by Duration of TZD Use and Discontinuation
| Women | Men | |
|---|---|---|
| n | 2371 | 4494 |
| Participants with fracture, n | 287 | 262 |
| Duration of TZD use | ||
| No TZD use | 1.00 reference | 1.00 reference |
| TZD use ≤1 y | 1.84 (1.17, 2.89) | 1.61 (1.03, 2.52) |
| TZD use >1 to 2 y | 2.32 (1.49, 3.62) | 1.44 (0.92, 2.27) |
| TZD use >2 y | 2.01 (1.35, 2.98) | 1.18 (0.79, 1.76) |
| Overall P valueb | <.01 | .17 |
| Time since last TZD use | ||
| Current TZD user | 1.00 reference | 1.00 reference |
| Discontinued TZD ≤1 y | 0.69 (0.46, 1.02) | 1.20 (0.81, 1.77) |
| Discontinued TZD >1 to 2 y | 0.57 (0.35, 0.92) | 1.12 (0.70, 1.80) |
| Discontinued TZD >2 y | 0.42 (0.24, 0.74) | 0.78 (0.43, 1.40) |
| Overall P valuec | <.01 | .53 |
Data are expressed as HR (95% CI), unless stated otherwise.
Adjusted for baseline variables: age, race/ethnicity, BMI, smoking status, diabetes duration, A1C, insulin use, TZD use, history of congestive heart failure, glycemia intervention randomization assignment, assignments to BP/lipid trials and interventions, and the other variables in the table.
Test of the null hypothesis that HR does not differ across categories of TZD use.
Test of the null hypothesis that HR does not differ across categories of TZD discontinuation.
Figure 1.
Net effects of TZD use and discontinuation, compared with never users, during ACCORD on non-spine fracture in women.
In secondary analyses, TZD use in women was associated with increased fractures of the lower extremities (P = .03) and marginally with upper extremity (P = .07) and axial fractures (P = .12) (Table 3). In men, none of these relationships were statistically significant. The effects of TZD discontinuation on fracture subtypes in women were directionally similar to those for all fractures, but only the association with lower extremity fractures (P = .01) was statistically significant. Because incident vertebral fractures were not available in ACCORD BONE, we considered the effects of TZD use on height loss, a surrogate marker for vertebral fracture. There were no differences in height loss across categories of duration of TZD use in women (P = .102) or men (P = .928). Height loss also did not differ across categories of TZD discontinuation in women (P = .487) or men (P = .170).
Table 3.
Adjusteda HRs for Non-Spine Fractures of the Upper Extremity, Lower Extremity, and Axial Skeleton by Duration of TZD Use and Discontinuation
| Women | Men | |
|---|---|---|
| n | 2371 | 4494 |
| Upper extremity fracturesb | ||
| Participants with fracture, n | 106 | 76 |
| Duration of TZD use | ||
| No TZD use | 1.00 reference | 1.00 reference |
| TZD use ≤1 y | 1.46 (0.69, 3.09) | 1.64 (0.71, 3.79) |
| TZD use >1 to 2 y | 2.05 (1.00, 4.19) | 1.69 (0.73, 3.89) |
| TZD use >2 y | 2.26 (1.19, 4.30) | 1.40 (0.65, 3.01) |
| Overall P valuec | .07 | .59 |
| Time since last TZD use | ||
| Current TZD user | 1.00 reference | 1.00 reference |
| Discontinued TZD ≤1 y | 0.49 (0.23, 1.03) | 1.09 (0.52, 2.31) |
| Discontinued TZD >1 to 2 y | 0.72 (0.34, 1.51) | 1.09 (0.44, 2.70) |
| Discontinued TZD >2 y | 0.44 (0.16, 1.18) | 1.32 (0.48, 3.58) |
| Overall P valued | .15 | .96 |
| Lower extremity fracturese | ||
| Participants with fracture, n | 175 | 134 |
| Duration of TZD use | ||
| No TZD use | 1.00 reference | 1.00 reference |
| TZD use ≤1 y | 2.00 (1.11, 3.61) | 1.67 (0.89, 3.12) |
| TZD use >1 to 2 y | 2.28 (1.29, 4.04) | 1.49 (0.79, 2.80) |
| TZD use >2 y | 1.41 (0.85, 2.33) | 1.26 (0.72, 2.18) |
| Overall P valuec | .03 | .42 |
| Time since last TZD use | ||
| Current TZD use | 1.00 reference | 1.00 reference |
| Discontinued TZD ≤1 y | 0.85 (0.53, 1.36) | 1.06 (0.61, 1.85) |
| Discontinued TZD >1 to 2 y | 0.48 (0.26, 0.91) | 1.19 (0.63, 2.23) |
| Discontinued TZD >2 y | 0.34 (0.16, 0.71) | 0.70 (0.30, 1.63) |
| Overall P valued | .01 | .72 |
| Axial fracturesf | ||
| No. of participants with fracture | 31 | 72 |
| Duration of TZD use | ||
| No TZD use | 1.00 reference | 1.00 reference |
| TZD use ≤1 y | 1.24 (0.30, 5.11) | 0.97 (0.40, 2.37) |
| TZD use >1–2 y | 1.98 (0.47, 8.33) | 1.06 (0.45, 2.54) |
| TZD use >2 y | 4.86 (1.25, 18.83) | 0.69 (0.33, 1.45) |
| Overall P valuec | .12 | .69 |
| Time since last TZD use | ||
| Current TZD use | 1.00 reference | 1.00 reference |
| Discontinued TZD ≤1 y | 0.63 (0.18, 2.20) | 1.52 (0.74, 3.09) |
| Discontinued TZD >1 to 2 y | 0.89 (0.24, 3.24) | 0.88 (0.33, 2.37) |
| Discontinued TZD >2 y | 0.83 (0.15, 4.53) | 0.75 (0.25, 2.28) |
| Overall P valued | .91 | .55 |
Data are expressed as HR (95% CI) unless stated otherwise.
Adjusted for baseline variables: age, race/ethnicity, BMI, smoking status, diabetes duration, A1C, insulin use, TZD use, history of congestive heart failure, glycemia intervention randomization assignment, assignments to BP/lipid trials and interventions, and the other variables in the table.
Upper extremity fractures included proximal humerus, upper arm, elbow and lower arm, distal forearm, wrist, hand, and finger.
Test of the null hypothesis that HR does not differ across categories of TZD use.
Test of the null hypothesis that HR does not differ across categories of TZD discontinuation.
Lower extremity fractures include hip, pelvis, upper leg, patella, lower leg, ankle, foot, heel, and toe.
Axial fractures include rib, chest/sternum, tailbone, and face.
Results for rosiglitazone use and discontinuation were similar to those for TZD use. In models adjusted for baseline variables (included in models for TZDs) and for time-varying pioglitazone use, the fracture rate was higher in women with 1–2 years of rosiglitazone use (HR = 2.16; 95% CI, 1.39, 3.36) or > 2 years of rosiglitazone use (HR = 1.96; 95% CI, 1.33, 2.88), compared with no use (Supplemental Table 1). The fracture rate was reduced in women who had discontinued rosiglitazone use for 1–2 years (HR = 0.48; 95% CI, 0.30, 0.79) or > 2 years (HR = 0.42; 95% CI, 0.24, 0.73) compared with current users. In men, the associations between rosiglitazone use or discontinuation and fracture rate were not statistically significant.
Discussion
We found that TZD use was associated with increased fracture risk in women, but not in men. We also found that attenuation of the increased fracture risk in women occurred after TZD use is discontinued. Our results for the effects of TZD use are consistent with results from post hoc analyses of randomized trials, showing an increased fracture rate for women with use of rosiglitazone or pioglitazone (3, 4). In ACCORD, TZD use was associated with about a doubling of the fracture rate compared with no use, similar to the pooled odds ratio of 1.9 reported in a meta-analysis of results from 22 randomized trials (4). We did not find increased risk with TZD use in men, also similar to pooled trial results (3, 4).
The reason for increased risk in women but not in men remains unclear. Some observational studies have reported that fracture risk with TZD use is similarly increased in men and women (5, 6). The usual upper age limit for enrollment in the randomized trials of TZDs was 75 years (12, 15), and the upper age limit for ACCORD enrollment was 79 years. In contrast, the case-control study nested in the UK General Practice Research Database included participants 80 years and older (5). Thus, it is possible that TZDs do increase fracture risk in men, but only at older ages. The distribution of fractures in men compared with women in ACCORD included a higher proportion of rib fractures, a fracture site that has a larger trabecular component. We considered whether this distribution might account for the difference in TZD effect by examining the results for fractures of the extremities separately from those of the axial skeleton. However, for TZD use, we found that the pattern of relative rates of fracture for the axial skeleton, primarily rib fractures in our analyses, was similar to that observed for the extremities. Women tended to experience higher fracture rates with TZD use, whereas there was little evidence of an effect of TZD use on fractures in men. Our results suggest that the higher proportion of rib fractures among men does not account for the observed gender difference. TZDs inhibit the aromatase pathway, reducing estrogen synthesis (16, 17). In older adults, men may be relatively protected from this effect by their higher levels of estrogen, compared with postmenopausal women. Although menopausal status was not ascertained in ACCORD, 95.3% of women were 55 years or older at baseline and presumably menopausal.
We report for the first time the effects of TZD discontinuation on fracture. Women who discontinued a TZD had a reduced fracture rate compared with current users. Our results indicate that the risk of fracture after discontinuation gradually returns to that of a nonuser, suggesting that the impact of TZDs on the skeleton is not permanent. Although the underlying mechanism is unclear, a similar pattern has been reported with cessation of glucocorticoid therapy, known to increase fracture rates. In a longitudinal study using data from the UK General Practice Research Database, the increased rate of vertebral and nonvertebral fractures returned to normal levels within 1 year after discontinuing oral glucocorticoids (18).
Randomized controlled trials have reported increased bone loss with rosiglitazone use (8, 19–21). Pioglitazone increased bone loss in premenopausal women with polycystic ovary syndrome (22) and in diabetic women and men (14), but it did not increase bone loss in women with impaired fasting glucose (23, 24). A rodent study (7) and clinical trial (8) have reported that the effects of rosiglitazone on bone loss are attenuated after discontinuation. In the trial comparing rosiglitazone and metformin, postmenopausal women assigned to rosiglitazone lost 1.7% more BMD at the femoral neck after 1 year than women assigned to metformin (8). In that study, women randomized to rosiglitazone were switched to open-label metformin and followed for an additional 6 months after the 12-month randomized comparison. After this additional 6 months, the rate of BMD loss was similar in the two groups, although BMD levels remained lower in women originally randomized to rosiglitazone.
We chose to combine rosiglitazone and pioglitazone use during ACCORD based on previous reports that identified increased fracture risk with both of these TZDs (4, 12, 25, 26). However, pioglitazone use was relatively infrequent in ACCORD, and we were not able to assess from our data whether the effects of TZD use or discontinuation varied between pioglitazone and rosiglitazone.
There are several other important limitations to this study. Our follow-up only included an average of about 3 years of use, followed by about 3 years of discontinuation. Thus, the results may not apply to longer durations of TZD use. Although ACCORD was a clinical trial, this analysis of TZD use is observational and therefore subject to confounding from unmeasured or poorly measured factors that are associated with TZD use and fracture. When participants discontinued TZD use, they may have initiated a different diabetes medication. In ACCORD, among those discontinuing rosiglitazone who did not initiate pioglitazone, the most frequently added medication was acarbose (W. Ambrosius, personal communication). These analyses did not control for use of other diabetes medications during the trial, and we cannot disentangle the effects of TZD discontinuation from the effects of starting another medication. Assessment of TZD use did not include dose. In addition, the number of pills dispensed to participants was recorded, but a pill count for compliance was not obtained at the visits. As a result, participants with the same duration of TZD use may have used different amounts of the medication. Given the longitudinal design, this potential misclassification is not likely to differ with respect to the outcome of fracture and would therefore tend to bias any real associations toward a null result. Fractures were initially reported by participants and may have been under- or over-reported. Fractures were centrally adjudicated based on medical records, eliminating nearly all instances of over-reporting. Thus, fracture identification was highly specific. In the setting of high specificity of the outcome, lower sensitivity will reduce power but will not bias the relative risk estimates in a longitudinal study (27).
In conclusion, TZD use in the ACCORD trial was associated with increased fracture risk in women, but not men, with type 2 diabetes who are at increased risk of CVD. When women discontinued TZD use, the fracture effects were gradually attenuated, indicating that these effects are reversible.
Acknowledgments
The authors thank Lisa Palermo, MS, for assistance with data management.
The ACCORD BONE ancillary study was funded by a grant (R01DK069514) from the National Institute of Diabetes and Digestive and Kidney Diseases. The ACCORD study was supported by grants (N01-HC-95178, N01-HC-95179, N01- HC-95180, N01-HC-95181, N01-HC-95182, N01-HC-95183, N01- HC-95184, IAA-Y1-HC-9035, and IAA-Y1-HC-1010) from the National Heart, Lung, and Blood Institute; by other components of the National Institutes of Health, including the National Institute of Diabetes and Digestive and Kidney Diseases, the National Institute on Aging, and the National Eye Institute; by the Centers for Disease Control and Prevention; and by General Clinical Research Centers. The following companies provided study medications, equipment, or supplies: Abbott Laboratories, Amylin Pharmaceutical, AstraZeneca, Bayer HealthCare, Closer Healthcare, GlaxoSmithKline, King Pharmaceuticals, Merck, Novartis, Novo Nordisk, Omron Healthcare, Sanofi-Aventis, and Schering-Plough.
Parts of this study were presented in abstract form at the annual meeting of the American Society for Bone and Mineral Research in Baltimore, Maryland, in October 2013.
ACCORD BONE Trial Registration: Clinicaltrials.gov NCT00324350; https://clinicaltrials.gov/ct2/show/NCT00324350?term=nct00324350&rank=1.
Disclosure Summary: A.V.S. received grant support from GlaxoSmithKline (outside the submitted work) and lecture fees from Merck and Chugai Pharmaceutical. A.S. received grant support from Novo Nordisk (outside the submitted work). R.G.J. received lecture fees from Merck, Lilly, Janssen, and Novo Nordisk and served on medical advisory boards for Merck, Lilly, Janssen, and AstraZeneca/Bristol-Myers Squibb. M.A.B. received grant support from Merck and Takeda and personal fees from Merck, Sanofi and Novartis. R.M.C. received grant support from Sanofi and Bristol-Myers Squibb (both outside the submitted work). T.I. received personal fees from Bayer. The other authors have nothing to disclose.
Footnotes
- A1C
- hemoglobin A1C
- BMD
- bone mineral density
- BMI
- body mass index
- BP
- blood pressure
- CI
- confidence interval
- CVD
- cardiovascular disease
- HR
- hazard ratio
- TZD
- thiazolidinedione.
References
- 1. Cauley JA. Public health impact of osteoporosis. J Gerontol A Biol Sci Med Sci. 2013;68(10):1243–1251. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. 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(4):427–444. [DOI] [PubMed] [Google Scholar]
- 3. Loke YK, Singh S, Furberg CD. Long-term use of thiazolidinediones and fractures in type 2 diabetes: a meta-analysis. CMAJ. 2009;180(1):32–39. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Zhu ZN, Jiang YF, Ding T. Risk of fracture with thiazolidinediones: an updated meta-analysis of randomized clinical trials. Bone. 2014;68:115–123. [DOI] [PubMed] [Google Scholar]
- 5. Meier C, Kraenzlin ME, Bodmer M, Jick SS, Jick H, Meier CR. Use of thiazolidinediones and fracture risk. Arch Intern Med. 2008;168(8):820–825. [DOI] [PubMed] [Google Scholar]
- 6. Douglas IJ, Evans SJ, Pocock S, Smeeth L. The risk of fractures associated with thiazolidinediones: a self-controlled case-series study. PLoS Med. 2009;6(9):e1000154. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Kumar S, Hoffman SJ, Samadfam R, et al. The effect of rosiglitazone on bone mass and fragility is reversible and can be attenuated with alendronate. J Bone Miner Res. 2013;28(7):1653–1665. [DOI] [PubMed] [Google Scholar]
- 8. Bilezikian JP, Josse RG, Eastell R, et al. Rosiglitazone decreases bone mineral density and increases bone turnover in postmenopausal women with type 2 diabetes mellitus. J Clin Endocrinol Metab. 2013;98(4):1519–1528. [DOI] [PubMed] [Google Scholar]
- 9. Schwartz AV, Margolis KL, Sellmeyer DE, et al. Intensive glycemic control is not associated with fractures or falls in the ACCORD randomized trial. Diabetes Care. 2012;35(7):1525–1531. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Buse JB, Bigger JT, Byington RP, et al. Action to Control Cardiovascular Risk in Diabetes (ACCORD) trial: design and methods. Am J Cardiol. 2007;99(12A):21i–33i. [DOI] [PubMed] [Google Scholar]
- 11. Gerstein HC, Miller ME, Byington RP, et al. Effects of intensive glucose lowering in type 2 diabetes. N Engl J Med. 2008;358(24):2545–2559. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Kahn SE, Zinman B, Lachin JM, et al. Rosiglitazone-associated fractures in type 2 diabetes: an analysis from A Diabetes Outcome Progression Trial (ADOPT). Diabetes Care. 2008;31:845–851. [DOI] [PubMed] [Google Scholar]
- 13. Hamann C, Kirschner S, Günther KP, Hofbauer LC. Bone, sweet bone–osteoporotic fractures in diabetes mellitus. Nat Rev Endocrinol. 2012;8(5):297–305. [DOI] [PubMed] [Google Scholar]
- 14. Vittinghoff E, Glidden DV, Shiboski SC, McCulloch CE. Regression methods in biostatistics: linear, logistic, survival, and repeated measures models. 2nd ed San Francisco, CA: Springer; 2012. [Google Scholar]
- 15. Home PD, Pocock SJ, Beck-Nielsen H, et al. Rosiglitazone evaluated for cardiovascular outcomes in oral agent combination therapy for type 2 diabetes (RECORD): a multicentre, randomised, open-label trial. Lancet. 2009;373(9681):2125–2135. [DOI] [PubMed] [Google Scholar]
- 16. Rubin GL, Zhao Y, Kalus AM, Simpson ER. Peroxisome proliferator-activated receptor γ ligands inhibit estrogen biosynthesis in human breast adipose tissue: possible implications for breast cancer therapy. Cancer Res. 2000;60(6):1604–1608. [PubMed] [Google Scholar]
- 17. Seto-Young D, Avtanski D, Parikh G, et al. Rosiglitazone and pioglitazone inhibit estrogen synthesis in human granulosa cells by interfering with androgen binding to aromatase. Horm Metab Res. 2011;43(4):250–256. [DOI] [PubMed] [Google Scholar]
- 18. Van Staa TP, Leufkens HG, Abenhaim L, Zhang B, Cooper C. Use of oral corticosteroids and risk of fractures. J Bone Miner Res. 2000;15(6):993–1000. [DOI] [PubMed] [Google Scholar]
- 19. Grey A, Bolland M, Gamble G, et al. The peroxisome proliferator-activated receptor-γ agonist rosiglitazone decreases bone formation and bone mineral density in healthy postmenopausal women: a randomized, controlled trial. J Clin Endocrinol Metab. 2007;92(4):1305–1310. [DOI] [PubMed] [Google Scholar]
- 20. Berberoglu Z, Yazici AC, Demirag NG. Effects of rosiglitazone on bone mineral density and remodelling parameters in postmenopausal diabetic women: a 2-year follow-up study. Clin Endocrinol (Oxf). 2010;73(3):305–312. [DOI] [PubMed] [Google Scholar]
- 21. Borges JL, Bilezikian JP, Jones-Leone AR, et al. A randomized, parallel group, double-blind, multicentre study comparing the efficacy and safety of Avandamet (rosiglitazone/metformin) and metformin on long-term glycaemic control and bone mineral density after 80 weeks of treatment in drug-naïve type 2 diabetes mellitus patients. Diabetes Obes Metab. 2011;13(11):1036–1046. [DOI] [PubMed] [Google Scholar]
- 22. Glintborg D, Andersen M, Hagen C, Heickendorff L, Hermann AP. Association of pioglitazone treatment with decreased bone mineral density in obese premenopausal patients with polycystic ovary syndrome: a randomized, placebo-controlled trial. J Clin Endocrinol Metab. 2008;93(5):1696–1701. [DOI] [PubMed] [Google Scholar]
- 23. Bone HG, Lindsay R, McClung MR, Perez AT, Raanan MG, Spanheimer RG. Effects of pioglitazone on bone in postmenopausal women with impaired fasting glucose or impaired glucose tolerance: a randomized, double-blind, placebo-controlled study. J Clin Endocrinol Metab. 2013;98(12):4691–4701. [DOI] [PubMed] [Google Scholar]
- 24. Grey A, Bolland M, Fenwick S, et al. The skeletal effects of pioglitazone in type 2 diabetes or impaired glucose tolerance: a randomized controlled trial. Eur J Endocrinol. 2014;170(2):255–262. [DOI] [PubMed] [Google Scholar]
- 25. Dormandy J, Bhattacharya M, van Troostenburg de Bruyn AR. Safety and tolerability of pioglitazone in high-risk patients with type 2 diabetes: an overview of data from PROactive. Drug Saf. 2009;32(3):187–202. [DOI] [PubMed] [Google Scholar]
- 26. Nissen SE, Nicholls SJ, Wolski K, et al. Comparison of pioglitazone vs glimepiride on progression of coronary atherosclerosis in patients with type 2 diabetes: the PERISCOPE randomized controlled trial. JAMA. 2008;299(13):1561–1573. [DOI] [PubMed] [Google Scholar]
- 27. White E. The effect of misclassification of disease status in follow-up studies: implications for selecting disease classification criteria. Am J Epidemiol. 1986;124(5):816–825. [DOI] [PubMed] [Google Scholar]
