Date Published: September 29, 2009
Publisher: Public Library of Science
Author(s): Ian J. Douglas, Stephen J. Evans, Stuart Pocock, Liam Smeeth, Anthony Keech
Abstract: Ian Douglas and colleagues analyze records from the UK General Practice Research Database, and find that among individuals prescribed thiazolidinediones who develop a fracture, fractures are more common during periods of thiazolidinedione exposure than unexposed periods.
Partial Text: The use of thiazolidinedione antidiabetic agents has become widespread in the treatment of type 2 diabetes since their introduction in the late 1990s, although recently the safety of this class of medicines has been called into question –. Although most attention has focused on the possible vascular effects of thiazolidinediones, the ADOPT trial of rosiglitazone also detected a 2-fold increased risk of fractures among patients treated with rosiglitazone . Similar effects were seen in a pooled analysis of pioglitazone trials , suggesting a possible class effect. Surprising—and unexplained—features of the observed increased fracture risk in clinical trials were that the risk appeared to mainly involve fractures of the arm, hand, wrist, or foot, and that the increased risk was restricted to women. However, the individual clinical trials did not have adequate statistical power to reliably assess fracture risk. People prescribed and not prescribed thiazolidinediones are likely to differ in ways that are difficult to measure and control for, making most observational studies subject to confounding and difficult to interpret. We therefore applied the self-controlled case-series design (a within-person approach) to assess the risk of fracture associated with thiazolidinedione use. This approach eliminates fixed (non-time-varying) between-person confounding, which is not the case with alternative case-control or cohort designs . The study was based on primary care computerized clinical records from the United Kingdom-based General Practice Research Database (GPRD).
We identified 1,819 patients in the GPRD prescribed at least one thiazolidinedione and with a record of at least one fracture and background details for these patients are shown in Table 1. 1,356 patients received only prescriptions for rosiglitazone, 389 received only pioglitazone, and the remaining 74 received prescriptions for both rosiglitazone and pioglitazone at different times. 990 (54%) were female and the mean age at first thiazolidinedione exposure was 57.9 y for men and 65.4 y for women. The mean duration of unexposed follow up prior to thiazolidinedione use was 9.5 y, and the mean duration of exposure to a thiazolidinedione was 2.3 y. Arm, foot, wrist, or hand fractures were recorded for 905 patients, hip fractures for 150 patients, and spine fractures in 66 patients. The remaining 733 patients had fractures at other sites. Patients with hip and spine fractures tended to be older at first thiazolidinedione prescription: mean age of 70.5 y and 65.6 y for patients with hip and spine fractures, respectively. Regarding multiple fractures, 283 (16%), 64 (4%), and 25 (1%) had two, three, and four or more fractures, respectively.
In these data we have found an increased risk of fracture in users of thiazolidinediones. Our results suggest that the increased risk applies to both men and women and is observed at a wide range of fracture sites. In addition, we were able to look at the effects of treatment duration on fracture risk, and our results suggest the risk may increase steadily with exposure duration up to 7 y. The risk estimate with pioglitazone was comparable to that seen with rosiglitazone at each fracture site, although the results for pioglitazone failed to reach statistical significance given that relatively fewer patients registered in the GPRD receive pioglitazone. While a clear mechanism for a deleterious effect of thiazolidinediones on bone has not been identified, a recent review has highlighted a number of preclinical studies in which rosiglitazone was associated with an increase in bone loss .