Re: Pioglitazone use and risk of bladder cancer: population based cohort study
We thank Drs. Gallagher, Winocour, Ryder, Blann, and Davidson for their interest in our work. As noted by Gallagher and Winocour, the previous pioglitazone trials were designed and powered to assess efficacy and not safety endpoints. This is particularly true for endpoints such as bladder cancer, which are both rare and have long latency periods. Given these limitations of trials regarding the assessment of such safety outcomes, the presence of an imbalance between the randomized groups necessitates further enquiry. With respect to the Insulin Resistance Intervention after Stroke (IRIS) trial, there was indeed an imbalance in the number of bladder cancer events in the in the pioglitazone versus placebo groups (12 events (0.6%) vs 8 events (0.4%), respectively) (1). Although this imbalance did not achieve statistical significance, its importance is strengthened by that also observed in the PROactive trial (2). Importantly, while we acknowledge that some early events were subsequently excluded from the bladder cancer analysis of PROactive and agree with Gallagher and Winocour that such events are unlikely to be causally associated with the recently initiated treatment, the imbalance in bladder cancer events remained after the exclusion of these early events (6 vs 3) (2). With the presence of this imbalance in bladder cancer in the two largest and longest pioglitazone trials published to date, we believe that these trial data are both consistent with our observed results (3) and should raise some concerns regarding the safety of pioglitazone with respect to bladder cancer.
With respect to the IRIS trial (1), it is important to note that this trial differed from the PROactive trial (2) in that it randomized non-diabetic patients (and thus involved a population at a lower risk of bladder cancer), and that there was a concerted effort (for ethical and safety reasons) to exclude patients with history of or at high risk of bladder cancer. Moreover, more patients in the pioglitazone group than in the placebo group were “removed for safety reasons” after randomization, which included the presence “of or risk factors for bladder cancer” (146 vs. 117 patients, respectively) (1). Unfortunately, it is unclear if the reported number of bladder cancer events includes patients who developed bladder cancer after they have been removed from the study.
Drs. Gallagher, Winocour, Ryder and Blann are also concerned that our findings are due to residual confounding. However, these concerns ignore the robust pharmacoepidemiologic approach employed that specifically aimed at circumventing the methodological limitations of the previous studies. This approach included the use of a new-user design, a time-dependent exposure definition, lagging exposure to deal with latency and detection bias, marginal structural modeling to deal with potential time-dependent confounding, and a head-to-head comparison of pioglitazone versus rosiglitazone with adjustment for high-dimensional propensity scores. All of these approaches generated results that were consistent with those of our primary analysis. While the observational nature of our study is still subject to some residual confounding from an unknown and/or unmeasured confounder, we show in Supplementary Figure 3 that the observed association would only be fully explained by a very strong confounder (i.e., one strongly associated with both the exposure and outcome); it is unclear if such a hypothetical confounder actually exists, suggesting that residual confounding is an unlikely explanation for our observed results.
One of the key findings of our study is the absence of an association with rosiglitazone, a thiazolidinedione introduced the same year and for the same target population as pioglitazone. However, Drs. Ryder and Blann claim that it is not possible to assume that users of pioglitazone and rosiglitazone were similar; this is based on their hypothesis of preferential prescribing of pioglitazone to patients with “more advanced” disease given the cardiovascular concerns that were raised with rosiglitazone. We are not aware of any studies that have documented this preferential prescribing pattern in the UK or elsewhere. Furthermore, this hypothesis is not supported by our data. Indeed, as shown in Supplementary Table 5, pioglitazone and rosiglitazone users were remarkably similar on nearly all characteristics, including Charlson comorbidity score (a variable that captures several cardiovascular conditions).
Dr. Davidson is concerned that our percentage of bladder cancer events was high (5.86%) among patients exposed to pioglitazone compared to what has been reported in other observational studies (<0.50%). Unfortunately, Dr. Davidson’s calculation was incorrect as it was based on the number of patients who were exposed to pioglitazone at cohort entry only (921 patients) rather than the total number of patients who were exposed during the entire study period (10,951 patients). When using the correct denominator (n=10,951), the percentage of bladder cancer events in pioglitazone users drops to 0.49% (i.e., 54/10,951). That being said, the incidence rate and not percentage should be the metric on which studies are compared, as the latter ignores the duration of follow-up.
Finally, Drs. Gallagher, Winocour, Ryder, and Blann discuss the relatively low cost of pioglitazone and its potential benefits on other outcomes, either along or when used in combination with other antidiabetic drugs. Our study was not designed to examine the cost-effectiveness of pioglitazone or its effects on these other endpoints. Rather, it was designed to determine the bladder cancer risk of pioglitazone, and we believe that it has provided strong evidence that pioglitazone is associated with an increased risk of bladder cancer. This evidence now must be considered when weighing the potential benefits and harms of pioglitazone relative to other currently available antidiabetic drugs.
Ultimately, the pioglitazone-bladder cancer controversy will not be resolved with a randomized controlled trial as it would be unethical to design and conduct a trial for a safety issue such as bladder cancer. This is where well-conducted observational studies, together with safety signals generated by randomized controlled trials, can provide critical information to relevant stakeholders.
References
1. Kernan WN, Viscoli CM, Furie KL, Young LH, Inzucchi SE, Gorman M, Guarino PD, Lovejoy AM, Peduzzi PN, Conwit R, Brass LM, Schwartz GG, Adams HP, Jr., Berger L, Carolei A, Clark W, Coull B, Ford GA, Kleindorfer D, O'Leary JR, Parsons MW, Ringleb P, Sen S, Spence JD, Tanne D, Wang D, Winder TR: Pioglitazone after Ischemic Stroke or Transient Ischemic Attack. N Engl J Med 2016;374:1321-1331
2. Dormandy JA, Charbonnel B, Eckland DJ, Erdmann E, Massi-Benedetti M, Moules IK, Skene AM, Tan MH, Lefebvre PJ, Murray GD, Standl E, Wilcox RG, Wilhelmsen L, Betteridge J, Birkeland K, Golay A, Heine RJ, Koranyi L, Laakso M, Mokan M, Norkus A, Pirags V, Podar T, Scheen A, Scherbaum W, Schernthaner G, Schmitz O, Skrha J, Smith U, Taton J: Secondary prevention of macrovascular events in patients with type 2 diabetes in the PROactive Study (PROspective pioglitAzone Clinical Trial In macroVascular Events): a randomised controlled trial. Lancet 2005;366:1279-1289
3. Tuccori M, Filion KB, Yin H, Yu OH, Platt RW, Azoulay L: Pioglitazone use and risk of bladder cancer: population based cohort study. BMJ 2016;352:i1541
Rapid Response:
Re: Pioglitazone use and risk of bladder cancer: population based cohort study
We thank Drs. Gallagher, Winocour, Ryder, Blann, and Davidson for their interest in our work. As noted by Gallagher and Winocour, the previous pioglitazone trials were designed and powered to assess efficacy and not safety endpoints. This is particularly true for endpoints such as bladder cancer, which are both rare and have long latency periods. Given these limitations of trials regarding the assessment of such safety outcomes, the presence of an imbalance between the randomized groups necessitates further enquiry. With respect to the Insulin Resistance Intervention after Stroke (IRIS) trial, there was indeed an imbalance in the number of bladder cancer events in the in the pioglitazone versus placebo groups (12 events (0.6%) vs 8 events (0.4%), respectively) (1). Although this imbalance did not achieve statistical significance, its importance is strengthened by that also observed in the PROactive trial (2). Importantly, while we acknowledge that some early events were subsequently excluded from the bladder cancer analysis of PROactive and agree with Gallagher and Winocour that such events are unlikely to be causally associated with the recently initiated treatment, the imbalance in bladder cancer events remained after the exclusion of these early events (6 vs 3) (2). With the presence of this imbalance in bladder cancer in the two largest and longest pioglitazone trials published to date, we believe that these trial data are both consistent with our observed results (3) and should raise some concerns regarding the safety of pioglitazone with respect to bladder cancer.
With respect to the IRIS trial (1), it is important to note that this trial differed from the PROactive trial (2) in that it randomized non-diabetic patients (and thus involved a population at a lower risk of bladder cancer), and that there was a concerted effort (for ethical and safety reasons) to exclude patients with history of or at high risk of bladder cancer. Moreover, more patients in the pioglitazone group than in the placebo group were “removed for safety reasons” after randomization, which included the presence “of or risk factors for bladder cancer” (146 vs. 117 patients, respectively) (1). Unfortunately, it is unclear if the reported number of bladder cancer events includes patients who developed bladder cancer after they have been removed from the study.
Drs. Gallagher, Winocour, Ryder and Blann are also concerned that our findings are due to residual confounding. However, these concerns ignore the robust pharmacoepidemiologic approach employed that specifically aimed at circumventing the methodological limitations of the previous studies. This approach included the use of a new-user design, a time-dependent exposure definition, lagging exposure to deal with latency and detection bias, marginal structural modeling to deal with potential time-dependent confounding, and a head-to-head comparison of pioglitazone versus rosiglitazone with adjustment for high-dimensional propensity scores. All of these approaches generated results that were consistent with those of our primary analysis. While the observational nature of our study is still subject to some residual confounding from an unknown and/or unmeasured confounder, we show in Supplementary Figure 3 that the observed association would only be fully explained by a very strong confounder (i.e., one strongly associated with both the exposure and outcome); it is unclear if such a hypothetical confounder actually exists, suggesting that residual confounding is an unlikely explanation for our observed results.
One of the key findings of our study is the absence of an association with rosiglitazone, a thiazolidinedione introduced the same year and for the same target population as pioglitazone. However, Drs. Ryder and Blann claim that it is not possible to assume that users of pioglitazone and rosiglitazone were similar; this is based on their hypothesis of preferential prescribing of pioglitazone to patients with “more advanced” disease given the cardiovascular concerns that were raised with rosiglitazone. We are not aware of any studies that have documented this preferential prescribing pattern in the UK or elsewhere. Furthermore, this hypothesis is not supported by our data. Indeed, as shown in Supplementary Table 5, pioglitazone and rosiglitazone users were remarkably similar on nearly all characteristics, including Charlson comorbidity score (a variable that captures several cardiovascular conditions).
Dr. Davidson is concerned that our percentage of bladder cancer events was high (5.86%) among patients exposed to pioglitazone compared to what has been reported in other observational studies (<0.50%). Unfortunately, Dr. Davidson’s calculation was incorrect as it was based on the number of patients who were exposed to pioglitazone at cohort entry only (921 patients) rather than the total number of patients who were exposed during the entire study period (10,951 patients). When using the correct denominator (n=10,951), the percentage of bladder cancer events in pioglitazone users drops to 0.49% (i.e., 54/10,951). That being said, the incidence rate and not percentage should be the metric on which studies are compared, as the latter ignores the duration of follow-up.
Finally, Drs. Gallagher, Winocour, Ryder, and Blann discuss the relatively low cost of pioglitazone and its potential benefits on other outcomes, either along or when used in combination with other antidiabetic drugs. Our study was not designed to examine the cost-effectiveness of pioglitazone or its effects on these other endpoints. Rather, it was designed to determine the bladder cancer risk of pioglitazone, and we believe that it has provided strong evidence that pioglitazone is associated with an increased risk of bladder cancer. This evidence now must be considered when weighing the potential benefits and harms of pioglitazone relative to other currently available antidiabetic drugs.
Ultimately, the pioglitazone-bladder cancer controversy will not be resolved with a randomized controlled trial as it would be unethical to design and conduct a trial for a safety issue such as bladder cancer. This is where well-conducted observational studies, together with safety signals generated by randomized controlled trials, can provide critical information to relevant stakeholders.
References
1. Kernan WN, Viscoli CM, Furie KL, Young LH, Inzucchi SE, Gorman M, Guarino PD, Lovejoy AM, Peduzzi PN, Conwit R, Brass LM, Schwartz GG, Adams HP, Jr., Berger L, Carolei A, Clark W, Coull B, Ford GA, Kleindorfer D, O'Leary JR, Parsons MW, Ringleb P, Sen S, Spence JD, Tanne D, Wang D, Winder TR: Pioglitazone after Ischemic Stroke or Transient Ischemic Attack. N Engl J Med 2016;374:1321-1331
2. Dormandy JA, Charbonnel B, Eckland DJ, Erdmann E, Massi-Benedetti M, Moules IK, Skene AM, Tan MH, Lefebvre PJ, Murray GD, Standl E, Wilcox RG, Wilhelmsen L, Betteridge J, Birkeland K, Golay A, Heine RJ, Koranyi L, Laakso M, Mokan M, Norkus A, Pirags V, Podar T, Scheen A, Scherbaum W, Schernthaner G, Schmitz O, Skrha J, Smith U, Taton J: Secondary prevention of macrovascular events in patients with type 2 diabetes in the PROactive Study (PROspective pioglitAzone Clinical Trial In macroVascular Events): a randomised controlled trial. Lancet 2005;366:1279-1289
3. Tuccori M, Filion KB, Yin H, Yu OH, Platt RW, Azoulay L: Pioglitazone use and risk of bladder cancer: population based cohort study. BMJ 2016;352:i1541
Competing interests: No competing interests