The CLL11 clinical trial demonstrated the benefit afforded by GClb versus RClb in terms of efficacy.17According to the evaluation of the European Medicines Agency, the results obtained with GClb versus RClb are of such a magnitude that they evidence a clear clinical relevance in delaying disease progression and are of evident importance for the patients.29 This study has examined whether these clinical benefits can also be regarded as cost-effective. Two conclusions can be drawn from the results obtained. First, there is a significant gain in QALYs with GClb versus RClb. Specifically, the gain in QALYs for each patient treated with GClb instead of RClb would be 0.673. This difference would have a significant patient impact considering that a clinically relevant gain (the minimum QALY difference the patient is able to perceive) is generally estimated as 0.03 or 0.04 QALYs.30,31 Second, treatment with GClb versus RClb can be regarded as efficient considering the cost-effectiveness thresholds commonly used in Spain (€30,000–45,000).32
Still, the results of the study should be interpreted in light of several limitations. This study utilized a decision-analytic modeling approach, which by definition constitutes a simplified simulation of reality. The model utilized a Markov process, which is the standard method used in cost-effectiveness studies to represent the natural history in various diseases33,34 and the most widely used model for analyzing cost-effectiveness in CLL.35
The main limitation of the analyses was the lack of long-term data in CLL11.17 Although PFS data had reached the median at the time of the last data cut-off, the median OS was not reached for GClb and RClb arm.17 For this reason, OS was projected using longer-term postprogression survival data from CLL5 and CLL8 trials. The same postprogression survival was applied in both arms; this was considered a conservative approach. As a result, the OS Kaplan-Meier curve and the projected OS curve deviated from each other in both arms (Figure 2), which could be explained due to the uncertainty in the tail of the KM curve, population differences across the trials (CLL11, CLL5, CLL8), as well as differences in the progression definitions. Further data from CLL11 would be needed to confirm the tails in the model extrapolations.
It should be, however, acknowledged that the distribution selected to extrapolate PFS (Weibull) had the lowest incremental QALY gain for GClb, compared to the rest of the distributions used in sensitivity analyses (Table 3), while being a key driver of the model results.
Another aspect of the model that must be underscored refers to the price of obinutuzumab. According to the Spanish Royal Decree-Law 16/2012, medicinal products can be prescribed from the NHS according to a reimbursed price or from outside the NHS with the price reported by the marketing pharmaceutical company.36 The reimbursed price by the NHS, which is lower than the reported price, is confidential and is agreed by the MSSSI. For this reason, the economic model could not make use of this reimbursed price and had to resort to the reported price, which is the only published price.37Since the reimbursed price is lower than the reported (used in the base case) price, the ICER resulting from using the reimbursed price would be lower than that obtained in the base case. In order to try to overcome this limitation, univariate sensitivity analyses were performed, applying different obinutuzumab acquisition levels (10% and 20% lower than the reported price), assuming that the deduction agreed with the MSSSI in establishing the reimbursed price would fall within this interval.
It is important to highlight the fact that the efficacy of obinutuzumab and chlorambucil combination was obtained from an explanatory Phase III clinical trial.7 In addition, patients included in the pivotal clinical trial had very different comorbidities.7 Consequently, it would be of interest to undertake economic analyses of cost-effectiveness of the scheme evaluated for different comorbidities and in the real world.
Lastly, another limitation of our study refers to the utilities employed. These utilities were not obtained from the patients in the CLL11 trial, as would have been desirable, but were derived from a study involving 100 healthy individuals in the UK.21 Nevertheless, the methodology employed (time trade-off method) and the questionnaire used (EuroQol-5D) in this study have been widely used in similar economic models.38–40 As regards to the applicability of the utility data to the Spanish population, a European utility study (Spain included) using the EQ-5D questionnaire revealed greater variability among individuals than among countries,41 thus making the UK-derived estimates, to some extent, appealing to the Spanish setting too.