It’s that time of year where we’re busy heading into the final quarter of the year. In 2018, we have witnessed a fundamental change in the concept of value, going far beyond pharmacological properties and clinical effects, to not only to include patient value, quality of life and patient management, but also to integrate systems and societal value dimensions more broadly.Where are we in our journey towards the new frontier in the ‘paying for value’ discussion in the US?
Stage 2 Predict Outcomes within the Health Plan Population and Identify Key Risks
Therapies often perform differently in the real-life compared to the clinical trial setting. The phenomenon has been described as the “effi cacy-toeffectiveness” gap mentioned earlier. A commonly-held belief is that performance is always reduced in the real-life, but there are numerous examples of treatments performing better than in the RCT (e.g., long acting dopamine agonists in schizophrenia, anti-IgE mAb in asthma). There are several factors that can infl uence real-life effectiveness, and they can be grouped into two categories: drug use factors and patient population factors. Drug use factors include patterns of use, dose, duration of treatment, past history of exposure, co-prescription, and adherence to treatment. Patient population factors include age, gender, behaviors, comorbidities, disease stage and severity, genetic and risk factors relevant to the disease. Furthermore, these factors are in turn influenced by the health system (e.g., coverage/reimbursement, medical practice, or screening policies influence which patients receive a treatment and how it is used). The so called “Drivers of Effectiveness” (DoE) framework, which originated from Analytica Lasers’ participation in “GetReal”, part of the EU Innovative Medicines Initiative, is illustrated in Figure 18.
For real-life effectiveness to differ from clinical efficacy, the distribution of patients along some of these factors has to differ significantly in the real-life compared to the clinical trial setting and there needs to be a meaningful interaction between the factor(s) and the outcome. Based on our experience, there is usually a limited set of factors that truly influence effectiveness and the interacting effects are mostly universal. The relevant factors can be identified through previous studies reported in the literature, sometimes based on your own clinical data, or by setting up ad-hoc studies.
What needs to be remembered is that the interaction between these factors and the drug efficacy is actually the same – whether you are looking at pivotal clinical trials or the real world. While the drug properties don’t change from population to population, the distribution of risk factors and effect modifiers does. That is why drug effectiveness appears different in the OBA context. A clear understanding how this framework plays out in the relevant context enables robust estimates of effectiveness as we concentrate on exactly these interactions. Of course, the more data that can be integrated, both ‘on the drug’ and on informing the characteristic of the population in real life, the better.
Once the factors are identified and their effects understood, it becomes possible to build a model that will be able to predict the real-life effectiveness. This model can then be used for different countries and health plans, provided the distribution of the influencing factors is understood in the populations of interest. The key benefit of the predictive approach is that it enables the manufacturer to quantify the multiple sources of variance and uncertainty that will be encountered in the execution of an outcomes-based agreement, which are related to:
■ the difference between the real-life versus clinical trial setting discussed above
■ the inter-patient variability
■ the effect of time, including both the intrinsic evolution of the disease and outcome, as well as potential changes in how, or by whom, the drug is being used.
While any model in itself does not reduce the uncertainty (without which an OBA would be rather pointless), it is essential to understand source and magnitude in order to frame and provision the contract adequately and manage the risk properly. Once the model offers a solid foundation of your understanding on real world performance, it can be useful to set up a baseline pilot for a preliminary discussion of mutual perspectives, data management and financial implications (paid for under “Fair Market Value” fees for staff time and resource needs to analyze data). This would precede the execution of the full-fledged agreement and can enhance further stages of the implementation.