Re: The use of pioglitazone and the risk of bladder cancer in people with type 2 diabetes: nested case-control study
We thank Dr. Ryder for his interest in our paper and for raising the important issues of confounding and switching of high-risk patients to pioglitazone. As discussed in our paper (1), pioglitazone users were in fact different from non-thiazolidinedione (TZD) users. For this reason, we also included a comparison between patients ever exposed to pioglitazone and those ever exposed to rosiglitazone. Overall, these two TZD groups were similar for most characteristics (see Web Extra for data supplement). In addition, we adjusted for known bladder cancer risk factors, including smoking, obesity, and uncontrolled diabetes, in all of our regression analyses.
It is important to note that this statistical adjustment had a minimal effect on our point estimates (crude rate ratio: 1.87 versus adjusted rate ratio: 1.83), suggesting a minimal effect of confounding. Furthermore, given the large point estimates observed, a very strong confounder (i.e. one strongly associated with both the exposure and outcome) would be required to fully explain the observed association. It is unclear whether such a variable exists beyond those considered, although residual confounding is possible. We acknowledge this possible limitation in our paper. As for Dr. Ryder’s concern regarding the switching of high-risk patients from rosiglitazone to pioglitazone, we explicitly dealt with this possibility by creating an exposure group consisting of patients who were exposed to both TZDs. We observed no increased risk with this group, although only 2 cases and 56 controls were ever exposed to both TZDs.
In summary, residual confounding is a limitation of any observational study. With respect to our study question, we were able to adjust for known bladder cancer risk factors. Such adjustment did not materially affect the point estimates, suggesting that confounding had a limited role in the association. Finally, our results are supported by signals observed in randomized controlled trials, which were not subjected to the biases described above.
1. Azoulay L, Yin H, Filion KB, Assayag J, Majdan A, Pollak MN, Suissa S. The use of pioglitazone and the risk of bladder cancer in people with type 2 diabetes: nested case-control study. BMJ 2012: 30;344:e3645.
Rapid Response:
Re: The use of pioglitazone and the risk of bladder cancer in people with type 2 diabetes: nested case-control study
We thank Dr. Ryder for his interest in our paper and for raising the important issues of confounding and switching of high-risk patients to pioglitazone. As discussed in our paper (1), pioglitazone users were in fact different from non-thiazolidinedione (TZD) users. For this reason, we also included a comparison between patients ever exposed to pioglitazone and those ever exposed to rosiglitazone. Overall, these two TZD groups were similar for most characteristics (see Web Extra for data supplement). In addition, we adjusted for known bladder cancer risk factors, including smoking, obesity, and uncontrolled diabetes, in all of our regression analyses.
It is important to note that this statistical adjustment had a minimal effect on our point estimates (crude rate ratio: 1.87 versus adjusted rate ratio: 1.83), suggesting a minimal effect of confounding. Furthermore, given the large point estimates observed, a very strong confounder (i.e. one strongly associated with both the exposure and outcome) would be required to fully explain the observed association. It is unclear whether such a variable exists beyond those considered, although residual confounding is possible. We acknowledge this possible limitation in our paper. As for Dr. Ryder’s concern regarding the switching of high-risk patients from rosiglitazone to pioglitazone, we explicitly dealt with this possibility by creating an exposure group consisting of patients who were exposed to both TZDs. We observed no increased risk with this group, although only 2 cases and 56 controls were ever exposed to both TZDs.
In summary, residual confounding is a limitation of any observational study. With respect to our study question, we were able to adjust for known bladder cancer risk factors. Such adjustment did not materially affect the point estimates, suggesting that confounding had a limited role in the association. Finally, our results are supported by signals observed in randomized controlled trials, which were not subjected to the biases described above.
1. Azoulay L, Yin H, Filion KB, Assayag J, Majdan A, Pollak MN, Suissa S. The use of pioglitazone and the risk of bladder cancer in people with type 2 diabetes: nested case-control study. BMJ 2012: 30;344:e3645.
Competing interests: No competing interests