3.1 Drug Risk Issues and Advisories3.1.1 Advisories

Within the 40 concordant drug risk issues, a total of 386 individual advisories were issued over the 10 years (Table 1), with the MHRA issuing the most (123/308) and the TGA issuing the least (72/308). The FDA was the first of the four regulators to issue advisories about a particular harm for 47.5% (19/40) drug risk issues, with TGA 5/40 (12.5%) and HC 5/40 (12.5%) being the least likely to communicate first. The FDA used website alerts as its primary communication method (39/40, 97.5%), whereas the TGA mostly used bulletins or newsletters (31/40, 77.5%). The MHRA and HC used a mix of website alerts and DHPCs. All four regulators more frequently issued advisories on individual drugs (107/160, 66.9%) compared to a drug class or multiple drugs. Table 1 of the Electronic Supplementary Material provides an overview of the medicines (by Anatomical Therapeutic Chemical classification), and the number of associated risks listed in advisories.

Table 1 Overview of the 40 concordant drug risk issues and advisories3.2 Content Comparative Analysis

The last advisory issued by each regulator was included in the comparative content analysis (n = 160) for each drug risk issue.

3.2.1 Evidence Cited and Risk Quantification

The evidence of harm was cited in 130/160 advisories (81.3%), with the FDA and TGA providing data more frequently than the MHRA and HC (Table 2). No significant difference was seen between regulators in the frequency with which any form of evidence was cited (χ2 = 5.42, df = 3; p = 0.14). Thirty advisories did not provide evidence to support risk information, with the MHRA not citing evidence in 12/40 (30%) of advisories issued.

Table 2 Overview of how the four regulators reported on the evidence of harm

Adverse drug reaction reports (including literature case reports) were the most frequently referenced type of evidence overall 49/160 (37.7%). This was not equally seen across the regulators, with HC 17/32 (53%) and the TGA 15/35 (42.9%) referencing adverse drug reaction reports more often than the FDA 8/35 (22.9%) and MHRA 9/28 (32.1%). Meta-analysis and systematic reviews were considered the highest level of evidence in our hierarchy [22], and these were cited in 12 advisories (12/130, 9.2%). The FDA referenced observational or epidemiological studies more than the other regulators (FDA 7/35, 20%).

Risks were quantified in 98/160 (61.3%) advisories at a similar frequency across the four regulators. The most common type of quantification presented was the proportion of patients experiencing an adverse event 42/160 (42.9%) followed by count data (e.g. number of adverse drug reaction reports or people), which was included in 39.8% (39/160) of advisories. Quantification only within verbal risk descriptions (e.g. rare, uncommon) was more commonly seen in the advisories from the TGA 6/27 (22.2%).

3.2.2 Type of Advice Provided

Specific advice was given in 96/155 advisories (61.9%) directed at clinicians, though no significant difference was found between regulators (χ2 = 4.79, df = 3; p = 0.19), with the MHRA most often recommending some form of specific action to clinicians 29/40 (72.5%). The TGA, in contrast, provided specific advice in a smaller proportion of advisories 20/38 (52.6%). When advice was provided, the most frequent advice was to ‘stop’ or ‘do not prescribe’ to specific sets of patients (56/96, 58.3%), followed by monitoring and testing advice (54/96, 56.3%)

Approximately 54/155 (35%) of advisories provided general advice (Table 3). There was no significant difference between regulators in the use of general advice-only advisories (χ2 = 3.91, df = 3; p = 0.27). The most common statements in advisories only providing general advice were that the prescriber should be aware of the safety issue 31/54 (57.4%), followed by advising the clinician to follow the new or updated product information or label 16/54 (29.6%).

Table 3 Overview of how the four regulators provided advice to clinicians3.2.3 Deaths or Potentially Fatal Outcomes

For 14 of the 40 concordant drug risk issues (35.0%), at least one regulator communicated about deaths or potentially fatal outcomes. All four regulators communicated about deaths for eight (20.0%) of the drug risk issues (Table 2 of the Electronic Supplementary Materials).

3.3 Overall Similarities and Differences Between Regulators

No significant differences were seen between the regulators regarding which elements of content were included. Some variation was seen within each type of content, such as how the regulator chose to explain or justify issuing an advisory, how evidence of risk was presented and the type of advice provided to clinicians.

3.4 Case Study: Pioglitazone and Bladder Cancer

Pioglitazone and bladder cancer was identified as a case study from the concordant advisories based on the case study selection criteria, and the regulatory communications were obtained to explore, in detail, the extent of difference and similarity in communication between regulators.

3.4.1 Overview of Pioglitazone and Bladder Cancer Advisories

Within the 10-year SAFER database period, the four regulators issued 14 advisories [23,24,25,26,27,28,29,30,31,32,33,34,35,36] on pioglitazone and bladder cancer (Fig. 1). We also located one Australian DHPC [37] through documentation released in a disclosure log provided by the TGA under the Freedom of Information Act [38].

Fig. 1

Timeline of safety advisories on pioglitazone and bladder cancer issued by the US Food and Drug Administration (FDA), UK Healthcare products Regulatory Agency (MHRA), Health Canada (HC) and the Australian Therapeutic Goods Administration (TGA) between 2010 and 2016. *A direct healthcare professional communication (DHPC) was issued by the Australian sponsor at the request of the TGA, but the exact date is unclear

The MHRA [27, 29, 34, 35] and FDA [23, 24, 28, 36] issued four advisories each, and the TGA [26, 30, 33] and HC [25, 31, 32] issued three each. The TGA response to our FOI request confirmed that no subsequent DHPCs had been issued [39]. The Australian DHPC appears to have been released around the same time as the first alert, but the exact timing cannot be confirmed, as per an FOI response from the TGA [39].

3.4.2 Differences in Timing of Advisories

Figure 1 presents a timeline of the advisories concerning the association of pioglitazone and bladder cancer. The first advisory amongst the four regulators was issued in September 2010 by the FDA [23], notifying that a safety review was being undertaken on pioglitazone and the risk of bladder cancer. The following FDA alert in June 2011 [24] stated that an increased risk of bladder cancer was being added to the Warnings and Precautions Section of the drug label. Advisories issued by the TGA (on 18 July 2011 [26]) and MHRA (on 22 July [27]) indicated that updates to the Product Information or Summary of Product Characteristics were underway in consultation with the sponsor. The MHRA then posted the UK DHPC on 29 July 2011 [34] and a bulletin alert soon after (1 August 2011). In contrast, HC communicated in June 2011 that they were monitoring the potential risk but mentioned no additional action.

In October 2011, the TGA issued a Medicines Safety Update, which provided updated evidence [30]. Health Canada posted a DHPC [31] and a general audience alert [32] in April 2012, confirming that the Product Monograph was being updated. In December 2013, the TGA posted another Medicines Safety Update on the benefit-risk profile of pioglitazone [33]. The FDA issued the last advisory in the sample frame in December 2016 [36], presenting its conclusions on the increased risk of bladder cancer over 6 years from the initial alert.

3.4.3 Evidence Cited and Descriptions of the Risk of Bladder Cancer3.4.3.1 Overview and Evidence Cited as a Basis for Regulatory Action

All four regulators cited similar evidence (see Table 4), with three key studies indicating increased bladder cancer risks with pioglitazone primarily referenced (Table 3 of the Electronic Supplementary Material). First, the Kaiser Permanente Northern California (KPNC) study [40], a 10-year observational cohort (funded by Takeda, manufacturer of pioglitazone), had been requested by the FDA to further understand whether pioglitazone use increases the risk of bladder cancer. The second study was a retrospective cohort (Caisse Nationale d’Assurance Maladie des Travailleurs Salariés [CNAMTS]) conducted at the request of the French medicines’ regulator, using a similar methodology to the KPNC study [41].

Table 4 Matrix of evidence cited by each regulator on the association of pioglitazone use with bladder cancer

The third study (sponsored jointly by Takeda and Eli Lilly) was a randomised controlled trial, the PROspective pioglitAzone Clinical Trial In macroVascular Events (PROactive) [42]. Although this study was not explicitly designed to assess bladder cancer risks, more participants receiving pioglitazone were diagnosed with bladder cancer compared with placebo.

The FDA cited the unpublished KPNC 5-year interim analysis data as the basis for the initial advisory (Table 4 of the Electronic Supplementary Material). Results suggested a 40% increased risk associated with pioglitazone use of more than 12 months (hazard ratio [HR] 1.4, 95% confidence interval [CI] 0.9–2.1) and a statistically significant risk after 24 months of exposure [40]. The FDA also stated that evidence of bladder cancer risk was known from PROactive and the Liver Safety Study [43]. Additionally, there were pre-market carcinogenicity observations of bladder cancer in male rats already reported [23].

The Liver Safety Study, funded by Takeda, examined whether pioglitazone was associated with drug-induced liver injury. The published report does not mention any bladder cancer cases [43]. However, only liver-related serious adverse events and common adverse events occurring in > 7.5% of patients were reported. Based on the information provided in the HC advisories [31], we estimate that two patients taking pioglitazone and zero taking glyburide were diagnosed with bladder cancer in the Liver Safety Study.

All of the advisories issued by regulators in 2011 confirming a risk of bladder cancer referenced the KPNC study. CNAMTS was also cited by the FDA, MHRA and the TGA but not HC (Table 4 of the Electronic Supplementary Material). The CNMATS results indicated a statistically significant increased risk of bladder cancer compared with other antidiabetic agents (HR 1.22, 95% CI 1.03–1.43). Cumulative doses greater than 28,000 mg and exposures for 12 or more months showed an increased risk, generally adding to the likelihood of causation. The increased risk was seen in male, not female individuals [41].

The regulators differed in which studies they mentioned, with the PROactive study mentioned by all four regulators. Only HC and the FDA referenced the Liver Safety Study. Adverse drug reaction reports, an unpublished sponsor meta-analysis and epidemiological data were only mentioned by the MHRA [42]. This meta-analysis indicated that pioglitazone exposure was significantly associated with bladder cancer risk (HR 2.64, 95% CI 1.11–6.31) but became non-significant when patients exposed less than a year were removed [34, 35, 44].

In 2011, the FDA confirmed that updated prescribing information had been approved [28]. The MHRA reconfirmed the risk in a subsequent DHPC in January 2012 but clarified that pioglitazone should not be used for first-line treatment [35]. The TGA published a benefit-risk assessment of pioglitazone in 2013, outlining other known harms associated with pioglitazone and adverse drug reaction reports about bladder cancer (11 reports from June 2011 to September 2013) [33].

In 2016, the FDA issued an advisory stating they had concluded that an association existed between pioglitazone use and bladder cancer [36]. They noted conflicting evidence on the association, exposure and dose effects. No significant difference in risk was seen in the KPNC final results [45] (HR 1.06, 95% CI 0.89–1.26), with a weaker association seen in the 5-year analysis than in interim analyses. PROactive’s 10-year follow-up indicated that the increased risks found in the original PROactive clinical trial [46] did not appear to persist over time. A cohort study carried out in the Clinical Practice Research Datalink database (funded by the Canadian Institutes of Health Research) also found significantly increased risks of bladder cancer with pioglitazone use compared with no thiazolidinedione use (HR 1.63, 95% CI 1.22–2.19), with risks increasing with a longer duration and cumulative doses [47]. All four regulators noted that other regulators took action as a result of the KPNC and CNAMTS results. The TGA had the most information, specifically identifying where suspensions and safety messaging had occurred [26].

3.4.3.2 Risk Description

The four regulators agreed that there was an association between pioglitazone use and the risk of bladder cancer; however, this described risk varied (Table 4 of the Electronic Supplementary Material). The FDA [24] and the TGA [26] stated that the ‘use of pioglitazone for more than one year may be associated with an increased risk of bladder cancer’. Communications between the TGA and the Australian sponsor (Eli Lilly Australia Pty Ltd) document that the sponsor was given the opportunity to comment on the wording of the website alert, including the risk description before public release [48]. Eli Lilly Australia suggested a ‘potential’ risk of bladder cancer instead of the TGA language of an ‘increased’ risk of bladder cancer. The word ‘potential’ did not appear in the final website alert [26]. The HC DHPC and accompanying ‘public directed communication’ posted on the HC website, included the risk description ‘potential risk of bladder cancer’ and ‘potential increased risk of bladder cancer.’

The MHRA stated that there was a ‘small increased risk of bladder cancer associated with pioglitazone use.’ The length of exposure was not specified; instead, the focus was on the type of patient being prescribed the medicine—“it is not clear if the risk increases early in treatment or only after prolonged administration. However, with careful patient selection, the benefits of pioglitazone continue to outweigh the risks.” [27].

3.4.4 Advice Provided to Clinicians

The four regulators arguably all had access to the same evidence and data, yet their advice for clinicians differed. Table 5 summarises this advice.

Table 5 Advice provided to clinicians in the first advisory confirming the risk of bladder cancer by the regulators

All four regulators contraindicated use in active bladder cancer, but only the MHRA, HC and TGA contraindicated use in patients with a history of bladder cancer. The FDA only suggested caution in this situation. The MHRA [27] and HC [31] both contraindicated use in those with unexplained macroscopic haematuria. The MHRA further focused on the benefits of pioglitazone that remained for carefully selected patients, and avoiding use in older adults. Guidance was provided for ongoing monitoring and discontinuing patients who did not respond to treatment. Later, advice was provided that pioglitazone should only be considered as a second- or third-line therapeutic option.

The TGA took a different approach, stating that the risk of bladder cancer should be considered in all patients treated with pioglitazone [