The changes of rosiglitazone and pioglitazone utilisation were observed between 2004 and 2012 in Australia. It is always difficult to attribute cause to utilisation trends, however it is likely that increased marketing of TZDs may have contributed to the increasing trend of rosiglitazone during 2004–2006 or that fewer alternatives to metformin, sulfonylurea, and insulin were available at this time. Our results show a decreasing trend in rosiglitazone utilisation in the period after the drug authorities’ warnings in 2007–2008. Although the numbers of rosiglitazone prescriptions in Australia are relatively low in comparison to the UK, and North America, the overall trends are consistent with those shown in Europe and North America [24, 26, 27, 48]. There are two possible explanations for the dip seen in April 2007. It might be a seasonal trend as the same fluctuation was noted in March-April 2006; however, this was not sensitive enough to be detected by the ARIMA model. Secondly, the dip is an artifact of the data, this is actually the utilisation on its way up which is demonstrated by the higher use again in May 2007.
The sharply decreasing utilisation trend is significantly attributable to the safety alert from meta-analysis study and the initial warnings from the EMA and FDA in May 2007. For the reason that the FDA issued the cardiovascular alert of rosiglitazone on the same day as publication by Nissen et al. , we could not distinguish the effects between the authority warnings and the publication. Furthermore, the effects of these warnings and associated literature are likely to be cumulative rather than a discrete effect on the following utilisation. Several restrictions in rosiglitazone subsidies from the PBS during October 2008-Febraury 2009 were also examined; however, these impacts are not significant after adjustment for previous warnings. As a result of the consecutive series of cardiovascular warnings on rosiglitazone since 2007 and the limited access on PBS, the numbers of rosiglitazone prescriptions have remained lower than 5,000 per month since 2010.
Australian utilisation of pioglitazone was less than half of rosiglitazone during 2005–2007 and the increasing trend in use was moderate compared to the Netherlands and the US [25, 27]. From 2008–2010, when peak levels were reached, the increase in pioglitazone nearly mirrors the decline in rosiglitazone. The findings suggest that prescribers might have replaced rosiglitazone with the same drug class pioglitazone [24, 49], due to the reported cardiovascular benefits of pioglitazone, and no clinical outcome associated with an increase risk of ischemic heart disease that was seen with rosiglitazone [7, 17]. While the decreasing trend of pioglitazone was observed in the US and Europe in 2008 [27, 49], Australian pioglitazone utilisation plateaued until 2011. The delay in decreasing trend compared to that of other countries may be attribute to limited availability of second-line and third-line therapy alternatives such as sitagliptin (was not PBS subsidised until August 2008) or exenatide (was not PBS subsidised until August 2010) . The US and UK data [49, 51] show that the number of other new drugs, which were available in their markets since 2007 such as sitagliptin and exenatide, increased after the cardiovascular alerts of TZD. Nevertheless, Figure 1 shows the decline in the utilisation of pioglitazone after July 2011. This decreasing trend may have been caused by more alternative treatments on the PBS or the safety concern of increased risk of bladder cancer in long-term users of pioglitazone. Although, this decline was of a lesser magnitude than for rosiglitazone, prescribers may consider the risk/benefit ratio, where the benefits of pioglitazone in lowering blood sugar outweigh the possible risk of bladder cancer . However, more data points following this bladder cancer risk might be needed to examine the true effect of this warning.
Since TGA safety warnings are considered by the Australian Department of Health and Aging to be first-line alerts to Australian prescribers, we would expect to see a significant effect on these utilisations. However, the fact that a) the decline in rosiglitazone use occurred prior to the first TGA warning, and b) after we adjusted for the preceding EMA, FDA warnings, we could not see a significant effect of the TGA warning on utilisation trends suggests that Australian prescribers were aware of the international warnings as well as the safety information from the literature. This might be associated with the way that information was delivered, since Australian warnings were delayed, less frequently communicated, and accessed compared to the FDA and European warnings [38, 53]. Australian prescribers may receive safety information from medical articles or media that referred to the US or European warnings. A further qualitative study is being conduct to gain the insight into sources of drug safety information among Australian prescribers.
Since time series model prediction is based on the pattern of drug use in the past confounding influences on data may be difficult to disentangle. Although trends can be impacted by temporal changes in drug supply or the way data are recorded, we did not find those problems during study period. Furthermore, the Australia PBS data is aggregated data collected for administrative purposes, which does not link utilisation to the prescribing data in clinical settings. Therefore, clinical reasons for the decrease in dispensing cannot be fully investigated, nor primary non-compliance in patients be established.
The strength of this study is that it captures almost all prescriptions dispensed over 2004–2012 in total Australian population (private prescriptions represent a very small percentage of all prescriptions). This is achieved because rosiglitazone and pioglitazone are ‘high’ cost drugs that are government subsidised in Australia. This allowed us to investigate the patterns of population based thiazolidinedione utilisation.