Sequential drug decision problems in long-term medical conditions: a case Study of Primary Hypertension Eunju Kim ba, ma, msc
Yüklə 10,52 Mb.
|
2) Other parametersSensitivity analyses were undertaken to find out whether the optimal solution identified using enumeration in the base-case is sensitive when the objective of SDDP is to maximise the treatment success rate rather than total net benefit. In practice, costs might be of secondary concern to clinicians, who primarily seek to optimise clinical outcomes for individual patients. There was significant uncertainty in the time-dependent SBP lowering effect based on Wald’s and Wright et al’s systematic reviews. The assumptions in the base-case of the hypertension SDDP model is that the BBs and CCBs drop SBP quickly in the first three months and then have little incremental SBP lowering effect after three months, whereas Ds and ACEIs/ARBs gradually drop SBP over one year. In sensitivity analysis, all single antihypertensive drugs were assumed to decrease SBP gradually over one year. The long-term SBP lowering effect of Ds, CCBs and ACEIs/ARBs was derived from ALLHAT[317] and those of BBs was derived from Dutch-TIA[318]. In this case, the SBP lowering effect is highest in Ds, followed by CCBs, ACEIs/ARBs and BBs consistently over time. Three-monthly SBP change was calculated based on Equation 5.5. The same assumption with the base-case was used for the SBP lowering effect of both two and three-drug combinations (see Equation 5.6). The absolute SBP lowering effect was converted to the relative SBP lowering effect using Equation 5.7. Table 7.. Optimal treatment sequences and total net benefit depending on gender and initial age
1) The different levels of initial SBP were only applied to the base-case, which included a cohort of 60 year old male patients. 2) The policy numbers with * are included in the top eight policies identified from the enumeration in Table 7.. Figure 7.. SBP lowering effect calculated based on ALLHAT and Dutch-TIA
The uncertainties in AEs were also explored in sensitivity analyses. Whereas DAEs, which were used in the base-model, only include serious AEs that cause treatment discontinuation, AEs include any symptoms that might plausibly be caused by antihypertensive drugs. Law et al provided comprehensive information on the AEs of antihypertensive drugs[282, 319]. Furthermore, considering that the prevalence of AEs may be higher in a clinical setting than RCTs, it was expected that using AEs, which have higher rates of incidence than DAE, could provide more realistic results[281]. Lastly, sensitivity analysis was conducted to explore the uncertainties in treatment scenarios for CVD. In the base-case it was assumed that all patients who had a CV event or DM took a recommended antihypertensive drug to treat the underlying disease. In practice, however, it is possible to use another drug that is not recommended in the clinical guidelines, but is believed by the clinicians to be the best for the patients without contraindications (clinician’s opinion)[4]. In sensitivity analysis, patients, who had a history of a CV event or DM, were assumed to use a randomly selected drug. The exception was CCB for patients with HF as CCB is contraindicated in these patients[8-10, 245]. Table 7. summarises the results of the sensitivity analyses implemented depending on the change in key parameters. Sensitivity analyses showed that the optimal solution (or the policies, which are not significantly different with the optimal solution) identified in the base-case was robust to changes in objective, SBP lowering effect, the extension of drug switching period, AE rates and the treatment scenario for CVD and DM, apart from the fourth-line drug. Where the objective function was to maximise the treatment success rate during the drug switching period, the optimal solution started with ACEIs/ARBs, and then moved to Ds+ACEIs/ARBs, Ds+BBs+ACEIs/ARBs and Ds+BBs+CCBs. Where the time-dependent SBP lowering effects calculated based on ALLHAT and Dutch-TIA were used, the optimal solution started with ACEIs/ARBs, and then moved to Ds+ACEIs/ARBs, Ds+BBs+ACEIs/ARBs and Ds+CCBs+ACEIs/ARBs. Where DAE (instead of AE) or the random treatment scenario for CVD and DM (instead of the set of recommended drugs for CVDs and DM) was used, the optimal initial, second and third-line drugs were the same with where the time-dependent SBP lowering effects calculated based on ALLHAT and Dutch-TIA was used, but the optimal fourth-line was BBs+CCBs+ACEIs/ARBs. Where the drug switching period was extended to six or eight periods, Figure 7. shows the increase in the percentage of the patients who used all four lines of treatments in a defined sequential treatment policy. Compared with the base model, the percentage of the patients who used all four lines of treatments was increased between 0.021% and 74.85% above the rates estimated for the base-case (average 10.17%). Where the drug switching period was extended to six or eight periods, the total net benefits were also increased by £5,840 to £21,750 (average £9,305) compared with the base-case (see Figure 7.). However, the optimal solutions were the same with the optimal solution identified in enumeration, where the drug switching period was extended to eight, or the seven policies, which were not significantly different with the optimal solution, where the drug switching period was extended to six. The extension of the drug switching period increased the number of possible disease pathways from 31 to 127, where the drug switching period was six, and 511, where the drug switching period was eight. This also involved the additional modelling codes to allocate a drug based on the assumed decision rules and to allow calculating possible transitions and total net benefits during the extended periods. Due to the increase in the size of the problem and computational complexity, the computational time increased substantially from 12.20 hours in the based model to 73.65 hours, where the drug switching period was six, and to 222.15 hours, where the drug switching period was eight. Table 7.. Optimal solutions and total net benefits depending on the change in key parameters
1) The base-case is in grey. 2) The policy numbers with * are included in the top eight policies identified from the enumeration in Table 7.. Figure 7.. Proportion of patients who used four lines of treatments depending on the modelled drug switching period Figure 7.. Total net benefit depending on the modelled drug switching period
Yüklə 10,52 Mb. Dostları ilə paylaş:
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||