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

5.6.11Comparison between the hypertension SDDP model and the NICE hypertension model

The underlying evaluation model of the hypertension SDDP model was constructed based on the NICE hypertension model, which has been regularly updated since it was developed in 2006; therefore, both models share similarities in terms of the population’s characteristics, single treatment options and the structure of CVD model. However, the ultimate goals of the two models are very different. The objective of the NICE hypertension model is to identify the most cost-effective initial drug for the management of hypertension in primary care, whereas the objective of this hypertension SDDP model is to identify the optimal sequential treatment strategy in primary hypertension. Accordingly, the comparators in the NICE hypertension model are single antihypertensive drugs, whereas the comparators in the hypertension SDDP model are all possible combinations of single drugs and two or three-drug combinations. While the NICE hypertension model assumed that the initial drugs were used continuously over the follow-up period, the hypertension SDDP model allowed changing treatment regimen depending on the patient’s health state (i.e., the result of previous treatment) during the drug switching period.

Although the included health states are the same in both the NICE hypertension model and the hypertension SDDP model, the structures are different. While the NICE hypertension model is a conventional Markov model, the hypertension SDDP model has a successive decision tree with an add-on Markov model. The clinical drug switching rules were incorporated into the successive decision tree. To populate the clinical drug switching rules, the surrogate outcome modelling based on SBP was used. The CVD risk was calculated using QRISK2, which is a validated risk engine for the UK population. A shorter time cycle of three months, compared with the NICE hypertension model of six months, was selected to better reflect the time to revisit the clinician after treatment initiation and to decide whether the drug is well-responded and needs to be switched.

In addition to the novel underlying evaluation model, the hypertension SDDP model has an additional outer loop that inputs potential sequential treatment policies to the underlying evaluation model and assesses the optimality of the current policy based on the treatment net benefit. Due to the computational complexity and the size of the problem, Matlab (version 8.1, R2013a) was used to develop the model, whereas the NICE hypertension model was built by using Microsoft Excel.

Table ‎5.. Comparison between the SDDP hypertension model and the NICE hypertension model

	SDDP model in primary hypertension	NICE hypertension model^[63]
Objective	To identify the optimal treatment sequence, which maximises the total net benefit.	To identify the initial drug, which is the most cost-effective compared with competing major antihypertensive drugs.
Perspective	UK NHS perspective.	UK NHS perspective.
Time	Total follow-up period: Lifetime Cycle: 3months.	Total follow-up period: Lifetime Cycle: 6 months.
Population	Patients newly diagnosed with primary hypertension excluding those with pre-existing CVD, HF or DM.	Essential hypertension seen in primary care, excluding those with pre-existing CVD, HF or DM.
Cohort	Different cohorts, defined by age (50, 60 and 70), gender and initial level of SBP (153.5, 163.5 and 173.5 mmHg). Base-case population is 60-year old men with the SBP of 173.5 mmHg.	Different cohorts defined by age (55, 65, 75 and 85), gender, CVD risk (0.5-5% per year), HF risk (0-5% per year) and DM risk (0-5% per year). Base-case population is 65-year-old men and women with 2% CVD risk, 1% HF and 1.1% DM risk.
Modelling method	Successive decision tree with an added-on Markov model (built by Matlab).	Semi-Markov model (built by Excel).
Model structure	1. Success. 2. Failure. Uncontrolled SBP without CVD. UA. MI. Stroke. HF. DM. 3. Death.	Well. MI. HF. Stroke. UA. DM. Death.
Treatment options	4,128 combinations of: •  Ds. •  BBs. •  CCBs. •  ACEIs/ARBs.	•  Do nothing. •  Ds. •  BBs. •  CCBs. •  ACEIs/ARBs.
Treatment effectiveness	A surrogate outcome modelling based on SBP lowering effect.	The conventional RR approach.
Outcome	Total net benefit.	Cost-effectiveness and total net benefit.