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

3) Cluster analysis depending on the combination of initial and second-line drugs

Cluster analysis was also implemented to compare the average total net benefits depending on the combination of initial and second-line drugs. Full list of groups tested by ANOVA and Tukey’s HSD is provided in Table ‎7..

Figure ‎7. shows that there was a significant difference in total net benefit between the 36 groups defined by the initial and second-line drug (p<0.05). Regardless of the initial drug, using CCBs+ACEIs/ARBs as the second-line treatment provides higher total net benefits than other second-line treatment options: that is, Group 9 (£322,141 where Ds were used initially, followed by CCBs+ACEIs/ARBs), Group 18 (£321,945 where BBs were used initially, followed by CCBs+ACEIs/ARBs), Group 27 (£321,900 where CCBs were used initially, followed by CCBs+ACEIs/ARBs) and Group 36 (£322,009 where ACEIs/ARBs were used initially, and then adding CCBs to ACEIs/ARBs).

The average total net benefits of the policies using Ds+CCBs (Group 5, 14, 23 and 32) or Ds+ACEIs/ARBs (Group 6, 15, 24 and 33) as a second-line drug were also high regardless of the initial drug. These policies were significantly better than using a single drug (Group1, 2, 3, 10, 11, 12, 19, 20, 21, 28, 29 and 30) and using Ds+BBs (Group 4, 13, 22 and 31) or BBs+CCBs (Group 7, 16, 25 and 34) as the second-line drug.
Table ‎7.. The list of tested groups divided by the combination of initial and second-line drugs

	1st drug	2nd drug
Group 1	Ds	BBs
Group 2	Ds	CCBs
Group 3	Ds	ACEIs/ARBs
Group 4	Ds	Ds+BBs
Group 5	Ds	Ds+CCBs
Group 6	Ds	Ds+ACEIs/ARBs
Group 7	Ds	BBs+CCBs
Group 8	Ds	BBs+ACEIs/ARBs
Group 9	Ds	CCBs+ACEIs/ARBs
Group 10	BBs	Ds
Group 11	BBs	CCBs
Group 12	BBs	ACEIs/ARBs
Group 13	BBs	Ds+BBs
Group 14	BBs	Ds+CCBs
Group 15	BBs	Ds+ACEIs/ARBs
Group 16	BBs	BBs+CCBs
Group 17	BBs	BBs+ACEIs/ARBs
Group 18	BBs	CCBs+ACEIs/ARBs
Group 19	CCBs	Ds
Group 20	CCBs	BBs
Group 21	CCBs	ACEIs/ARBs
Group 22	CCBs	Ds+BBs
Group 23	CCBs	Ds+CCBs
Group 24	CCBs	Ds+ACEIs/ARBs
Group 25	CCBs	BBs+CCBs
Group 26	CCBs	BBs+ACEIs/ARBs
Group 27	CCBs	CCBs+ACEIs/ARBs
Group 28	ACEIs/ARBs	Ds
Group 29	ACEIs/ARBs	BBs
Group 30	ACEIs/ARBs	CCBs
Group 31	ACEIs/ARBs	Ds+BBs
Group 32	ACEIs/ARBs	Ds+CCBs
Group 33	ACEIs/ARBs	Ds+ACEIs/ARBs
Group 34	ACEIs/ARBs	BBs+CCBs
Group 35	ACEIs/ARBs	BBs+ACEIs/ARBs
Group 36	ACEIs/ARBs	CCBs+ACEIs/ARBs

1) The full list of tested groups divided by initial and second-line drugs is provided in Table ‎7.. 2) Cycle represents the mean of total net benefits; horizontal line represents the comparison intervals; the blue colour represents the group, which has the highest average total net benefit; the grey colour represents the group whose average total net benefit is not significantly different with the group, which has the highest average total net benefit; and the red colour represents the group whose average total net benefit is significantly different with the group, which has the highest average total net benefit. Figure ‎7.. Graphical summary of the mean and comparison intervals for each cluster classified by the combination of initial and second drugs

4) Cluster analysis of top 10 % solution

Cluster analyses were conducted to explore the characteristics of the top 10% of policy options regarding mean total net benefit. Figure ‎7. shows the distributions (%) of each drug or drug combination included in the top 10% policies, depending on the position of the drug or drug combination in the drug sequence. Of 413 top 10% policies, 27.1% had Ds as the initial drug, 26.9% were ACEIs/ARBs, 24.7% were CCBs and 21.3% were BBs. The second-line drugs distribution mainly comprised of CCBs+ACEIs/ARBs of 41.4%, Ds+ACEIs/ARBs of 40.9% and Ds+CCBs of 15.0%. For the third-line drug, three drug combinations such as Ds+CCBs+ACEIs/ARBs (32.0%), BBs+CCBs+ACEIs/ARBs (25.7%), Ds+BBs+CCBs (21.3%) and Ds+BBs+ACEIs/ARBs (18.4%) took a greater proportion than other drugs. The fourth-line drugs were more evenly distributed across 14 treatment options with Ds+CCBs+ACEIs/ARBs (8.96%) and Ds+ACEIs/ARBs (5.08%) being the highest and lowest.
Figure ‎7.. Distribution of the drugs included in the top 10% policies
A similar pattern was found in the second, third and fourth-line drugs where the top 10% of policy options were divided by initial drug (see Figure ‎7. - Figure ‎7.). For example, where a Ds was used initially (Figure ‎7.), the second-line drugs of the top 10% solutions mainly comprised of Ds+ACEIs/ARBs (42.7%) and CCBs+ACEIs/ARBs (42.7%). For the third-line drug, three drug combinations such as Ds+CCBs+ACEIs/ARBs (32.0%), BBs+CCBs+ACEIs/ARBs (22.3%), Ds+BBs+CCBs (21.4%) and Ds+BBs+ACEIs/ARBs (21.4%) took a greater proportion than other drugs. The fourth-line drugs were evenly distributed across the remaining 13 treatment options.
Figure ‎7.. Distribution of the drugs included in the top 10% policies, where Ds are used initially

Figure ‎7.. Distribution of the drugs included in the top 10% policies, where BBs are used initially

Figure ‎7.. Distribution of the drugs included in the top 10% policies, where CCBs are used initially

Figure ‎7.. Distribution of the drugs included in the top 10% policies, where ACEIs/ARBs are used initially

7.3.2Sensitivity analysis

1) Population characteristics

Sensitivity analyses were conducted to examine whether the base-case enumeration results were sensitive to gender, age and initial SBP. Table ‎7. summarises the optimal solutions and their net benefits depending on age, gender and initial SBP.

The total net benefit decreased as the age increased. It was higher in women than men across all age groups because women had a lower CVD risk so lived longer than men in the hypertension SDDP model. Despite the different baseline risks depending on the age and gender, most optimal treatment sequences were not out of the optimal solution identified in enumeration and the seven policies that were not statistically different with the optimal policy. The optimal initial drug was either ACEIs/ARBs or Ds across all age and gender groups. For both men and women whose initial treatment fails, optimal second-line treatment was either adding Ds to ACEIs/ARBs, where ACEIs/ARBs is used initially, or adding ACEIs/ARBs to Ds, where Ds is used initially. For the third-line treatment, Ds+CCBs+ACEIs/ARBs were optimal in most age and gender groups, which was identical with the NICE guidelines. For the fourth-line treatment, Ds+CCBs+ACEIs/ARBs or BBs+CCBs+ACEIs/ARBs were optimal, depending on the optimal drugs selected in the previous cycles. The computational time was longer in the patient group aged 50 initially (14.58 hours for men and 18.90 hours for women) than the patient group aged 70 initially (8.39 hours for men and 8.56 hours for women) due to how long the patients live.

The total net benefit was lower where the initial SBP was lower (i.e., 163.5 or 153.5 mmHg) than the base-case (173.5 mmHg). This result corresponds with MacMahon et al and Collins et al’ s studies, which showed the greater treatment benefit for an individual having a greater CVD risk at any age^{[224, 225]}. For the patients with a lower initial SBP than the base-case, the optimal initial drug was ACEIs/ARBs, which was same with the base-case. The optimal second-line drug was sensitive to the change in patients’ initial SBP. Where the initial SBP was 163.5 or 153.5 mmHg, the optimal second-line drug was CCBs+ACEIs/ARBs, whereas Ds+ACEIs/ARBs was the optimal second-line drug in other scenarios. This can be explained by the relative (rather than absolute) SBP lowering effect assumed in the hypertension SDDP model because, with the same relative effect, the difference in the absolute SBP reduction between Ds+ACEIs/ARBs and CCBs+ACEIs/ARBs was reduced if the initial SBP becomes lower. This could reduce the difference in total net benefit between the policies using Ds+ACEIs/ARBs and CCBs+ACEIs/ARBs as the second-line drug.