Real-world cost-effectiveness of cetuximab in the third-line treatment of metastatic colorectal cancer based on patient chart review in the Netherlands

Cancer is the most frequent cause of mortality worldwide, with an estimated 8.2 million cancer-related deaths in 2012. Colorectal cancer (CRC) is one of the most prevalent cancers, the third most commonly diagnosed cancer in men and the second in women, with 746,000 and 614,000 new cases annually, respectively. Additionally, it is the third-leading cause of cancer-related mortality worldwide, with 373,500 deaths annually [1].

The incidence of CRC is expected to increase in the coming decades. Therapies containing oxaliplatin, irinotecan, and biologics have improved overall survival (OS). In the mid-1970s, 5-year OS rates were 50%, and this rate has since increased to 66% [2]. Biologic agents targeting vascular endothelial growth factor, its receptor, or epidermal growth factor receptor (EGFR) are the mainstay of biologic treatment options.

Two drugs targeting EGFR have been licensed, cetuximab and panitumumab [3, 4]. For these drugs, reimbursement within the Dutch healthcare system was regulated under an “expensive drugs policy.” This policy allowed for conditional reimbursement, contingent on delivering evidence of real-world drug utilization and demonstrating associated cost-effectiveness in Dutch daily practice after 4 yrs. In this paper we present the results of a real-world study concerning cost-effectiveness of cetuximab in third-line treatment of patients with KRAS (exon 2) wild-type (wt) metastatic CRC (mCRC) compared with best supportive care (BSC).

To our knowledge, this study is the first of its kind to calculate a real-world ICER for cetuximab in this setting. The results led to an ICER of €133,527/QALY (95% CI, €92,521-€184,072/QALY).

Our study has some limitations. First, clinical records did not contain sufficient information to determine PFS in the BSC group and, furthermore, no quality-of-life data were available in the medical records. To overcome this shortcoming, data from randomized controlled trials (RCTs) were used to populate the model in cases in which real-world data were lacking. RCT data were extrapolated and combined with real-world data to populate the BSC arm of the model, and we also applied quality-of-life data derived from the same RCT [12]. In the RCTs and in real-world data, treatment with cetuximab compared with BSC was associated with almost a doubling of the median OS and PFS among patients with KRAS wt tumors.

In an RCT, patients are treated according to a predefined protocol and have more active follow-up than patients treated under real-world conditions. Additionally, RCTs often apply stricter inclusion and exclusion criteria (e.g., patients with good performance status and few comorbidities), whereas in the real world, older patients with a multitude of comorbidities and poorer condition are also eligible [13]. In the RCTs, more patients had colon and/or rectum cancer. In the real-world data, there were relatively more rectosigmoid cancers. Concerning the performance status, the availability of recorded data was too limited to draw a conclusion about differences between the states of patients. Further, the RCT patients in general received more adjuvant treatments than the real-world patients did. As a result, treatments given may have been less intensive and the outcome of these treatments may have been worse than those seen in RCTs [13].

Our study design did not allow for randomization of patients. However, by matching patients from the control group to patients in the cetuximab group, we corrected for differences between the two groups at baseline. This way, known possible confounders were corrected for. However, whether eligible patients may receive cetuximab in daily practice depends on physician and patient choice; the rationale is not always provided in the medical record. As a result, we could not directly correct for these possible confounders.

Further, we assumed similarity of the effect ratio between the clinical trial and routine daily practice. The literature has shown that when patient characteristics were similar between RCT data and real-world data, assuming similarity in effect size yielded a valid approximation of relative effectiveness outcomes [14]. In our study, these differences, were marginal and tested in the PSA.

Cost data collection was conducted from a hospital perspective rather than a societal perspective [11]. A hospital perspective was chosen for this study because a societal perspective was not feasible. Societal costs were therefore not collected; however, due to the patients’ extended survival, the main sources of nonmedical costs in the Netherlands are often productivity loss or additional medical expenses unrelated to the condition under study. The population in this study had a mean age of 63 years, while the mean age of retirement in the Netherlands in 2009 was < 63 years [15]. Productivity losses, or potential gains, will therefore be minimal in this population. Additionally, because the survival of patients with mCRC often does not extend beyond 2 years, even with optimal treatment, it is unlikely that medical expenses unrelated to the primary condition would yield a large difference between the treated and BSC groups in this study. However, travel costs are a type of cost that may differ between these two groups as patients treated with cetuximab require weekly hospital visits to receive treatment, whereas patients receiving only BSC will visit less regularly. The differences in costs may be limited because cetuximab is a treatment that can be given at nearly any hospital, and the average distance to a hospital in the Netherlands is only 7 km [16].

Additionally, home care provided to a patient by a spouse, family member, or friend may lead to loss of productivity for the caregiver, and both longer survival and side effects in patients receiving cetuximab may result in more loss of productivity for their caregivers than would occur for patients receiving BSC only. We did not estimate the impact of these costs on the ICER.

For the cost analysis, different sources were used. Some limitations may be introduced by drawing data from different sources; however, the rationale for using different sources is inherent to the study design. Some data could not be obtained from hospitals, such as the resource use of BSC. However, Dutch guidelines provide a valid and reliable framework for such resource use and associated costs. Furthermore, the impact of the different costs has been studied by conducting a PSA.

Scientific articles on the cost utility of cetuximab monotherapy are scarce [17]. Hoyle et al. [18] reported an ICER of €127,300/QALY for cetuximab compared with BSC based on an indirect treatment comparison. In that study, utility values were adapted based on general population values, and resource use was partially based on patients treated for breast cancer.

Tappenden et al. [19] performed a threshold analysis from a UK National Health Service perspective. The conclusion of this analysis was that it was unlikely that an ICER of < €40,200/QALY could be achieved with cetuximab, even when a rule was applied that would mandate treatment to be stopped if no effect was observable after 6 weeks. Mittmann et al. [20] calculated the cost effectiveness of cetuximab treatment in Canada and reported an ICER for cetuximab of €209,729/QALY compared with BSC, substantially higher than the result of €134,777/QALY found in this study. This difference may be due to resource use in the analysis by Mittmann et al. being determined through trial data rather than daily practice. Additionally, the Mittmann publication references a higher per-mg cost of cetuximab than the drug cost used in our calculations.

Publications on the cost-effectiveness of panitumumab, a biologic with a similar marketing license, are scarce as well [17]. Hoyle et al. [18] reported an ICER for panitumumab in the United Kingdom of €250,580/QALY compared with BSC. The only other report on panitumumab cost-effectiveness was published only as an abstract by Graham et al. [21]. The ICERs reported were for the comparison of panitumumab plus BSC vs. BSC alone as third-line treatment of patients with KRAS (exon 2) wt mCRC. The ICERs were €51,314/LY gained and €59,440/QALY gained. These estimates were based on RCT data only, and the analyses were performed in a Dutch setting. Because the report is in abstract form only, the information presented was limited [17].

An upper bound of €80,000/QALY has been suggested for the Netherlands for the most severe illnesses [22]. Recently, however, there has been discussion on whether drugs used during the last stages of illness (end-of-life products) should be held to a different standard, as is the case in the United Kingdom [23, 24]. If so, a higher threshold might be more appropriate for such drugs. Additionally, the panitumumab indication was changed based on an analysis published by Douillard et al. [25], whereas the cetuximab indication was changed based on the re-analysis of the OPUS and CRYSTAL studies with regard to NRAS and KRAS exons 3 and 4 mutations [26, 27].