Sequential drug decision problems in long-term medical conditions: a case Study of Primary Hypertension Eunju Kim ba, ma, msc

8.2.2Implications of Hypertension SDDP modelling for decision-makers

Despite extensive economic evaluations that have been performed on antihypertensive drugs, previous economic evaluations have been mostly focused on comparing the cost-effectiveness of initial treatment options Ds, BBs, CCBs, ACEIs and ARBs; however, the principles of evidence-based medicine should be also applied to drug switching because the selection of subsequent treatments is directly associated with treating a patient with hypertension effectively and managing healthcare expenditure efficiently in the long-term. The results of the hypertension SDDP model show that decision-makers may risk potentially accepting a locally optimal solution if subsequent treatments are not explicitly modelled.

The hypertension SDDP model is a comprehensive and flexible model, which could address a wide range of questions that would interest decision makers. The hypertension SDDP model is capable of not just analysing the cost-effectiveness of initial drugs (as can be seen from the validation scenario in Section 7.2), but also evaluating the impact on health outcomes and resource use for subsequent treatment options. The main goal of the hypertension SDDP model was to identify the optimal treatment pathway in primary hypertension. While the conventional economic evaluation involves the development of partial decision models to identify the cost effective treatment options for patients in a certain period of the entire disease pathway, the hypertension SDDP model is a single comprehensive modelling framework that is capable of evaluating the cost-effectiveness of multiple treatment options across the broader pharmacological treatment pathway.

In SDDPs, decision-makers may have different research questions depending on their perspective. Although the decision-makers generally pursue maximising total treatment net benefits in SDDPs, some decision-makers may be interested in developing evidence-based stepped-care guidelines from a population perspective (i.e., what is the optimal initial drug and then what is the optimal second or third-line drug after the previous drug fails to control the disease), and some decision-makers may be interested in dynamic treatment assignment from an individual patient’s perspective to provide a tailored optimal treatment sequence to individual patients’ need for treatment. Both perspectives can be addressed in the hypertension SDDP model. Like the approaches used for enumeration, SA and GA, the former population-perspective decision problem can be addressed by constructing the decision space with complete solutions (i.e., all possible treatment sequences) and then examining probabilistically the aggregated results of all individual patient’s cases in a cohort model. In the latter individual-perspective decision problem, RL would be the natural choice to facilitate the interactive sequential decision-making process, considering inter-patients variability in risk factors and response to treatment.

For movement to the comprehensive but complex modelling approach, Lord et al said that the comprehensive modelling approach could be able to produce consistent cost-effectiveness estimates under a common framework of methods, baseline data and assumptions, but the adaption of such a large and complex model by regulatory and HTA bodies is still uncertain because of the considerable resource requirements and time constraints^[39]. The hypertension SDDP model introduced potential methods to reduce the computational complexity and to save computational time for a large and complex problem. Enumeration may be of limited use in many large and complex SDDPs, especially where high speed computers are not available and efficient and flexible programming languages are not supported, because the computational times are likely to be too long to be of use in practice. In this situation, heuristic methods were proposed as a good alternative to find optimal or near optimal solutions in a reasonable amount of time. The proposed approaches are particularly useful in long-term medical conditions that have a lot of potential health states and treatment options.