Sequential drug decision problems in long-term medical conditions: a case Study of Primary Hypertension Eunju Kim ba, ma, msc
The structure of the long-term CVD model
Yüklə 10,52 Mb.
|
- Bu səhifədəki naviqasiya:
- 5.6Treatment effectiveness and costs
5.5The structure of the long-term CVD modelThe long-term CVD model is a standard Markov model, which calculates the long-term cost and effectiveness of sequential treatment strategies. The model structure with the following states is identical with the NICE hypertension model[63].
Like the NICE hypertension model, other CVDs such as stable angina, peripheral vascular disease and transient ischemic attacks were not included because data on them were not consistently reported in the trials[63]. During the drug switching period, a proportion of patients in both Success and Failure, who had CVD or DM, moved to the long-term CVD model. After the drug switching period, all patients alive moved to the long-term CVD model. The initial distribution of health states in the long-term model depended on the final treatment result in the drug-switching period. If a patient had a CVD or DM at the end of the drug switching period, for example, they were allocated to one of CVDs or DM in the long-term model. Otherwise, the patient started the long-term transitions from Well. A cohort of patients with hypertension moved through the long-term CVD model until death or 100-years old. Table 5. shows the possible transitions between the health states included in the long-term CVD model. Every year patients who had no experience of CVD and DM either stayed in the same state or moved to one of the CVD states, DM or death. Once a CV event happened, it was assumed that a proportion of patients has a recurrent event, moves to another CVD state or moves to the post-CVD states. Same assumption was applied to the post-CVD states. The transition from a post-CVD to Well was not allowed because the post-CVD states involved the CVD history. Patients in the DM state were assumed either to stay in the same state or to move to one of CVD states. Table 5.. Health states and possible transitions in the long-term CVD model
Baseline transition probabilities for the patients having CVD or DM also came from the NICE hypertension model (see Table 5.)[63]. Because the NICE hypertension model calculated the baseline transition probabilities based on a population with 2% CVD risk, those transition probabilities were adjusted to be relative to the patient's final CVD risk in the drug switching period. Those patients who have never achieved the treatment goal during the drug switching period were assumed to move to Well in the long-term CVD model after correcting the underlying cause. They were assumed as resistant hypertension, which was defined as someone whose blood pressure remains above 140/90 mmHg despite the optimal concurrent use of three antihypertensive agents of different classes[63]. The prognosis of resistant hypertension is unclear and the pharmacological treatment for resistant hypertension, which involves three or four-drug combinations, has not been systematically evaluated[294]. Thus there was a difficulty to consider these patients in the framework of the SDDP of primary hypertension. Instead, the average costs and utility decrement of CVDs were used for the negative impact of resistance hypertension on costs and effectiveness. For patients who have CVD or DM, antihypertensive treatment should be initiated to reduce their CVD risk together with surgical or pharmacological care to treat the underlying disease. There may be a recommended drug for patients with a (history of) specific health state. For example, ACEIs (or ARBs as an alternative) are generally advised for patients with chronic HF[295]. For patients with (a history of) ischemic heart disease, BBs or CCBs are acceptable. For a person on antihypertensive treatment at diagnosis of DM, NICE recommends to initiate pharmacological treatment with ACEIs (or ARBs for a person with continuing intolerance to ACEIs) or CCBs[63]. However, these are recommendation, but not strong compelling indications. In practice, it is possible to use another drug, which is not recommended in the clinical guidelines, but is believed to be the best for the patients without contraindications (Clinician’s opinion). In the base-case, it was assumed that all patients who had or currently have CVD or DM took a recommended antihypertensive drug to treat the underlying disease (i.e., CCBs for UA, BBs for MI and stroke and ACEIs for HR and DM), whereas they were assumed to use a randomly selected drug in a sensitivity analysis apart from CCBs for patients with HF. Table 5.. Annual baseline risks of primary and secondary CV events[63]
5.6Treatment effectiveness and costs5.6.1Surrogate outcome modelling based on systolic blood pressureSurrogate outcome is a laboratory measurement or a physical sign that is intended to substitute for a clinically meaningful patient outcome[296, 297]. The use of surrogate outcomes has the merit of smaller sample size, shorter follow-up period and less cost to dissemination of new treatments, whereas conducting a randomised clinical trial for the final outcomes can be expensive and take a long time. However, there are concerns about the use of surrogate outcomes because many studies showed that the use of surrogate outcome as a final outcome has been misleading for decision-making[298-300]. To validate the surrogate outcome, it must be in the causal pathway of the disease process and all effects of intervention on final outcome should be fully captured by a change in the surrogate outcome. For HTA, the aim of the use of surrogate outcomes is to predict a clinically important final outcome where data on the final outcomes are not available and the causal relationship between surrogate outcomes and final outcomes is well established[246, 286]. Taylor and Elston recommended that “Ideally, the assessment of clinical effectiveness and cost-effectiveness of a health technology should be based on final patient-related outcomes. When this is not possible, CE analysis based on a surrogate outcome can be considered where there is evidence demonstrating a pathophysiological and clinical consistent association between the surrogate outcome and final patient-related outcome and corresponding treatment effects between them”[297]. In the hypertension SDDP model, SBP was used as the surrogate outcome because of the nature of SDDP. In practice, SBP is an important clinical factor to decide a treatment regimen and to observe whether the treatment is successful or not; clinical guidelines also clearly state the target blood pressure. If a patient’s blood pressure is not reduced to this target blood pressure, a clinician should consider stopping the current drug safely and trying another treatment regimen[63]. There is abundant evidence of a positive and consistent relationship between blood pressure and CV events[224, 227, 301]. Some CEAs in primary hypertension used a risk engine to populate the CV events[258, 302, 303]. The existing risk engines include not just SBP but also various risk factors that are important to address the patient’s CVD risk. The most widely used Framingham equation was developed based on a large and long-term community-based cohort in the US[304]. The 10-year CVD risk is estimated based on age, gender, SBP, smoking, TC, HDL and DM. However, Cooper et al pointed out that Anderson’s Framingham equation model tends to overestimate the absolute risk of CVD in European populations including the UK, and several risk factors, such as family history, ethnic group, socio-economic status, hypertension treatment and extremes of risk factors, are not included[305]. Hippisley-Cox et al said that D'Agostino’s Framingham model, which is a newer Framingham model (see Table 5.), uses a much broader definition of CVD that is less relevant to UK guidelines[285]. As an alternative, QRISK is a new risk score that has been developed using routine data from UK electronic primary care patient records[285, 306]. It has the advantage that practices and patients on the database are representative of the UK population; hence, QRISK reduces the uncertainty arising from generalising between the variation in population and practice. QRISK also includes ethnicity, BMI, family history of CHD, Townsend deprivation score, treated hypertension, CRD, AF and RA, which are not included in the Framingham equation. The validation of QRISK1 showed a better discriminator of 10-year CVD risk in the UK population compared with the Framingham risk score and the newly developed Scottish score (ASSIGN)[306]. A revised equation for QRISK (QRISK2) was more predictive of CVD risk in the UK cohort compared with the Framingham risk score and the initial version of QRISK[285]. This hypertension SDDP model applies the open sources of QRISK2 to consider the relationship between SBP and CVD risk[307]. Table 5.. Comparison between Framingham models and QRISK2
Yüklə 10,52 Mb. Dostları ilə paylaş:
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||