Abiraterone for castration-resistant prostate cancer: adherence, survival and hospitalization

In the past 15 years oral anticancer medications have become more widely available for the treatment of a variety of cancers and have increasingly been used as an alternative to intravenous (iv) therapy. This trend is also demonstrable in the management of advanced prostate cancer (PC). The introduction of the oral drugs abiraterone acetate (AA) and enzalutamide [1] has revolutionized the management of castration resistant prostate cancer (CRPC). The availability of oral anticancer drugs aligns with the preferences of cancer patients and may help to improve patient quality of life and reduce the time spent in healthcare settings; however, the availability of oral drugs induces the issue of drug compliance or adherence [2, 3]. Both terms define the extent to which patients take medications as prescribed by their healthcare providers. Without adequate adherence, the efficacy of oral drugs might fall below of those administered intravenously. The second threat of an oral drug, particularly if well tolerated such as AA and enzalutamide, is that these drugs might be prescribed too liberally.

Prompted by the paucity of data of both aspects we investigated these issues by analyzing the database of the largest public health insurance in Austria (Wiener Gebietskrankenkasse, WGKK). This database was compared to the Austria death statistics and hospital admission registry. For decades Austria has had a public healthcare system with compulsory state insurance. The high prevalence of CRPC as well as the cost of these drugs further emphasize the economic impact of drug adherence and prescription pattern.

After receiving institutional board approval from our hospital the prescription database of the largest public insurance company in Austria the WGKK was reviewed. Data from all patients with at least one prescription of AA from November 2011 to December 2014 were extracted, where AA [1] is typically prescribed at 1000 mg daily taken orally with 10 mg oral prednisone. The following data were extracted from this database: age, data of the first AA prescription, number of AA prescriptions, number of hospital admissions and duration of hospital admissions. This database was matched to the Austria death and hospital admission statistics.

Adherence [2, 3] was calculated using the medication possession ratio (MPR), which is the sum of all days of AA supplied within a given period, divided by the total number of days in that period. There is no consensus standard for what constitutes adequate adherence. Some trials consider rates of greater than 80 % to be acceptable [4, 5].

The principal findings of this Austrian prescription database analysis were (i) high adherence to AA, (ii) high mortality within the first months of AA prescription, (iii) 24 % long-term survivors under AA and (iv) high rate of hospitalization in patients under AA.

Prior to discussion, several pros and cons of our approach need to be discussed. Strengths are: (i) the population-based character of this database. The WGKK is the biggest insurance company in Austria (8.4 million insured persons) and it has 1.2 million members in all socioeconomic classes, (ii) for decades Austria has had a public and equal access healthcare system with compulsory state health insurance company and (iii) long follow-up and complete survival data due to matching with the Austrian death statistics. The main limitation is the lack of any clinical information (e.g. indications, previous chemotherapy, staging and reason for drug discontinuation). Finally, the data on hospital admissions have to be interpreted in the context of the Austrian healthcare system.

Cancer treatment is evolving. Chronic oral administration transfers responsibility from the practitioner to the patient, making adherence an important parameter in reducing the risk of treatment failure. Adherence to medical treatment is a complex and multifaceted issue that can substantially alter the outcome of therapy [6‐10].

Outside randomized controlled trials (RCT) few studies have reported on the adherence to AA. Smith et al. [4] analyzed pharmacy claims of the Canadian Saskatchewan Cancer Agency. All patients with at least one AA prescription were eligible and a total of 86 patients were followed for a minimum of 6 months. Optimal drug adherence was achieved in 82.6 % of patients with 79.1 % reaching a MPR of at least 90 %. At 6 months the mean MPR was 89.6 % (median 100 %) and after 12 months 86.6 % (median 99.5 %). Lafeuille et al. [5] studied this issue by analyzing two large-scale US administrative healthcare claims databases. The mean age of the patients was 72.2 years and the mean MPR was 93 % (median 98 %). The mean MPR in our series was 94.8 ± 11.9 (median 100 %) with no relevant impact of patient age in contrary to the study by Grundmark et al. [6]. Taken together, these two studies with a substantial number of analyzed patients indicate that the adherence to AA in a real-life setting is satisfactory with median values exceeding 95 %.

In contrast to the prescription database studies described previously, this current one is unique with respect to the availability of survival data. There is a major concern that survival data generated by RCT or registries do not reflect the real life setting [11‐16]. It has been previously shown that eligibility to a chemotherapy protocol represents per se a good prognostic factor [17]. In the pivotal phase III trial of AA in the postchemotherapy setting de Bono et al. [18] reported on a median survival of 14.8 months as compared to 11 months in our series. The median age of our patients, however, was 6 years older compared to the de Bono et al. [18] trial. The maximum follow-up in our study was considerable longer (38 months) as compared to the pivotal phase III trial with 20 months. There are a considerable number of long-term survivors. In our series, 24 % of patients survived longer than 20 months and 37 % in the phase III trial. The early mortality rate, however, was substantially higher in our series. Within the first 6 months of AA treatment 39 % of our patients died as compared to only 15 % in the de Bono et al. phase III trial [18]. These data suggest that patient selection in real life is substantially less stringent than in a phase III trial. In our series, patient age had a relevant impact on overall survival. In patients aged 46–76 years the median survival was 17 months and thus longer than in the pivotal RCT as compared to only 5 months in those older than 76 years. Houede et al. [19] reported on the long-term outcome of the AA temporary authorization for use program in France: 306 patients with a median age of 63 years were analyzed. The overall survival in this cohort after initiation of AA was 14.6 months, almost identical to the RCT and longer than in our series. As expected, OS was correlated to the duration of AA treatment, a similar phenomenon was observed in our series (see Fig. 3).

The third aspect of this study was the analysis of hospital admissions and length of hospitalization after initiation of AA. In our series, these patients spent a considerable time of their remaining life in hospital care. Only 7 % were not hospitalized and one quarter of patients (26.2 %) spent more than 50 % of their remaining life span in hospital care. These data have to be interpreted in the context of the Austrian healthcare system, where admission to hospital care is liberal and free of charge to the patient; furthermore, there is no incentive to discharge patients as early as possible. We could not identify a comparable analysis with CRPC patients under chemotherapy and second-line endocrine therapy in the literature.

According to Svensson et al. [20] patients were on average 2 years older than those in the RCT, which is partly in agreement with our study population, OS in Swedish patients was the same like de Bono et al. [18] (COU-AA-301 trial), in contrary to our results, and the researchers concluded that the treated population and treatment patterns, organization of healthcare, as well as country setting could contribute to differences in outcomes between the clinical trial and the real world treatment, which could be an explanation for our OS results in comparison to the randomized controlled trials.

Outcomes of treatment in clinical practice can differ from outcomes in RCT with regard to the estimated effectiveness and the estimated resource utilization. Discrepancies in patient and physician behavior between the trial and the real world may have an impact on outcome and treatment cost.

In clinical trials patients continued treatment until documented progression while the real world evidence study collected information on progression. There is much interest in confirming whether the efficacy of AA demonstrated within the trial setting is reproducible in routine clinical practice in a non-trial setting and many differences should be taken in consideration between both, such as the selection of patients and ethnic differences.

The OS in the study of Poon et al. [21] of chemotherapy-naive patients was 18.1 months and much shorter than that reported in the COU-AA-302 study (of 34.7 months) Ryan [16] and a higher proportion of elderly patients, which is in agreement with our study.

This Austrian prescription database allows some relevant insights into the outcome of patients treated with AA for CRPC in a real life setting. Drug adherence was satisfactory and OS was shorter as compared to the pivotal phase III trial. The high early mortality rate in our series suggests poor patient selection in real life. One quarter of patients experience long-term survival. The hospitalization rate within this cohort was substantial.