This paper has investigated, for the first time in the literature, the economic consequences of 11 CUP for medicines from the perspective of health care payers in Italy. Avoided costs of treating patients with the SoC, if existing and reimbursed, incremental costs of diagnostics, combination therapies, and side effects of CUT were included. Savings generated by the CUP ranged from €26.5 million (€10.9k per patient) to €50.5 million (€25.6k per patient), depending on the SoC used as an alternative to the CUT.
These findings cannot be fully compared with the economic impact of a CT. If Italian studies on CT are considered, including only medicines [7, 17] and diagnostic procedures , the mean savings per patient treated are higher in our findings. This is mainly caused by the higher cost of the hypothetical SoC used in clinical practice.
It is important that policy makers and health care managers appropriately and carefully consider this evidence.
First, the findings represent costs avoided from the perspective of health care payers and not of the society as a whole.
Secondly, incremental costs, despite quite negligible, are costs generated by CUP, whereas avoided costs are contingent upon the actual use of alternative treatments.
Thirdly, costs of medicines are variable and coincide with expenditure. Costs of diagnostic testing and inpatient care have been estimated using the respective fee-for-service/episode. Fees represent a cost/expenditure for payers, whereas, from the perspective of providers, they represent revenue and do not necessarily correspond to cost.
Finally, net savings from compassionate use should not justify delays in decision-making on P&R upon approval of the medicine, on the grounds of shifting the burden from the payer to the industry. In principle, any early access program, like compassionate use, should terminate when the drug is approved. However, delays in P&R decisions  have resulted in pharmaceutical companies prolonging CUP, to avoid interruptions in the continuity of care. Further, companies should not expect to secure reimbursement and/or a higher price for their medicines on the grounds of the latter’s subjection into a CUP. On the one hand, not all medicines used in CUP were approved for reimbursement in Italy. On the other hand, prices should reflect value  and not whether the relevant medicine has been granted an early access.
Our study has some limitations.
The first and most important one is that we have not estimated the future impact on costs of CUT compared to the SoC. CUP are not accompanied by systematic collection of data on effectiveness and resource consumption (apart from diagnostic testing and hospitalisations due to adverse events). An estimation of future net costs of the CUT would have required to adapt, through micro-simulation, published cost-effectiveness analyses on patients treated in CUP. Such a “dynamic” study was out of scope as our objective was to estimate the costs incurred throughout the duration of CUPs but not after. In addition, should we have adapted existing evidence from cost-effectiveness studies comparing the CUT to the SoC, we would be able to rely only on one study conducted in Italy .
Secondly, our analysis adopted the perspective of health care payers. To this end, the impact of CUP on other third payers (e.g., social insurance schemes for absence from work) and patients and their relatives (e.g., remunerated / informal care provided) was not considered. Indeed, for some diseases, like multiple sclerosis, costs beyond health care may play an important role.
Thirdly, since the comparative impact on disease progression and future cost avoidance were not included, we did not carry out an economic impact analysis where no alternative to the CUT was available. The scientific understanding of key molecular pathways and the development of new molecular entities in most of the diseases targeted by the examined CUP present opportunities to fill unmet needs, even in areas with already existing alternatives.
The absence of a long-term follow-up of patients recruited in CUP did not allow for the evaluation of the CUP effects on therapeutic sequencing. It may happen that the SoC is used after the CUT, or vice versa when patients are not recruited in the CUP (i.e., they are first treated with the SoC and after with the CUT, once it has been approved for reimbursement).
Furthermore, our analysis did not include an estimate of the cost of adverse events for the SoC, that should have been retrieved from the literature (unless it is used as a comparator in trials for the CUT). We are not able to conclude on whether this averted cost would have compensated the incremental cost of treating side effects of CUT that have not produced a hospitalisation, as this was not considered in the present analysis. We were also not always able to include the effect of discounts and MEA in the calculation of the cost of the SoC. Finally, the costs of nusinersen for Type-1 Spinal Muscular Atrophy refer to data from the first year of treatment, since real world data on treatment duration beyond the first year were not available.
Incremental costs for diagnostic procedures were estimated on the grounds of the presumed clinical practice with the SoC and assuming maximum impact for the health care payers. For example, we decided to include (i) IHC (Immuno-Histo-Chemistry) to detect Anaplastic Lymphoma Kinase (ALK)-positive, despite its use with crizotinib as an alternative to alectinib for adult patients with advanced NSCLC and (ii) FISH for trastuzumab emtansine, despite its with trastuzumab as a neo-adjuvant therapy. If the cost of these two diagnostic procedures were eliminated, the incremental cost would decrease from €4.5 million to €4.3 million with a net economic benefit ranging from €26.7 million (instead of €26.5) to €50.8 million (instead of €50.6 million).
Finally, while different SoC were identified, we did not have any information on their use in clinical practice. This information could have allowed for the calculation of a weighted mean cost per patient receiving the respective SoC. Thus, we are only able to provide a minimum and maximum average averted cost per patient, where the actual mean cost depends on the present market share of each single SoC.
Nevertheless, despite these limitations, this paper constitutes the first evidence on the economic advantages of the compassionate use of medicines.