An anchored simulated treatment comparison of uptitration of amlodipine compared with a low-dose combination treatment with amlodipine 5 mg/bisoprolol 5 mg for patients with hypertension suboptimally controlled by amlodipine 5 mg monotherapy

Figures & data

Table 1. Baseline characteristics and blood pressure outcomes of the studies that contributed data to the model.

	NCT00558428	BA5-A5 study
	Amlodipine 10 mg	Bisoprolol 5 mg and amlodipine 5 mg
Randomized Patients, No	276	181
Age, mean (SD), years	54.3 (10.6)	57.1 (13.7)
Female, n (%)	100 (36.2)	90 (49.7)
Race, n (%)
Caucasian	213 (77.2)	181 (100.0)
Asian	55 (19.9)	0 (0.0)
Black	6 (2.2)	0 (0.0)
Other	2 (0.8)	0 (0.0)
Body mass index, mean (SD) (kg/m^2)	28.6 (5.0)	29.6 (4.9)
Duration of hypertension (years)
< 1	80 (29.0)	43 (23.8)
1–5	88 (31.9)	68 (37.6)
6–10	52 (18.8)	29 (16.0)
> 10	56 (20.3)	41 (22.7)
Concomitant diabetes, n (%)	24 (8.7)	34 (18.8)
Smoking history, n (%)
Never smoker	175 (63.4)	83 (45.9)
Ex-smoker	61 (22.1)	35 (19.3)
Current smoker	40 (14.5)	31 (17.1)
missing	0 (0.0)	32 (17.7)
Blood pressure at baseline (i.e. post-amlodipine 5 mg run-in), mean (SD), mmHg
Systolic	149.3 (12.0)	148.8 (8.2)
Diastolic	96.5 (4.7)	90.2 (7.6)
Analyzed Patients, No	261	178
Change from baseline to 8 weeks, Least Squares Mean (standard error), adjusted for baseline blood pressure and country, in mmHg
Systolic	–11.1 (0.7)	–15.5 (1.0)
Diastolic	–8.0 (0.5)	–10.2 (0.7)

Abbreviations. SD, standard deviation. Sources: Neldam 2011, ClinicalTrial.gov.

Figure 1. Design of the model.

Abbreviations. AgD, aggregated data; BP, blood pressure; D1, difference in BP observed; D2, difference in BP predicted; IPD, individual patient data; RCT, randomized clinical trial.

Table 2. Fundamental steps in development of the statistical model.

Step	Methodology applied
	The Method was applied twice: once for SBP, and once for DBP
1	Development of a linear regression model
	Endpoint: BP at week 8 for both the treatment and the control arm was the outcome variable. Effect modifiers and their interactions with treatment with bisoprolol 5 mg/amlodipine 5 mg as well as prognostic variables were included as independent variables. Starting from a minimal model, including all relevant effect modifiers and interactions, prognostic variables were selected using forward selection. The best model was chosen by 5-fold cross-validation and the Akaike information criterion. The final model was fitted based on the results of the variable selection described above Graphical model-checking verified the linearity assumption and homogeneity of variance
2	Estimation of relative treatment effects between low-dose amlodipine/bisoprolol combination therapy and amlodipine 5 mg
	These models were then applied to predict SBP and DBP at week 8, for the combination and the amlodipine 5 mg arms, in a population comparable to that enrolled by the NCT00558428 study. The prediction utilized the mean covariate values (proportions for categorical variables) in the NCT00558428 study population. This permitted derivation, in each arm, of the BP decrease between baseline and week 8. This was followed by derivation of a predicted difference in BP decrease between the combination versus amlodipine 5 mg
3	Population-adjusted indirect comparison between low-dose amlodipine/bisoprolol combination therapy and amlodipine 10 mg
	A population-adjusted indirect comparison of the relative antihypertensive efficacy of amlodipine 5 mg/bisoprolol 5 mg combination therapy and amlodipine 10 mg monotherapy was made by considering the difference between the estimate obtained in Step 2 minus the observed decrease in BP between baseline and week 8 in the amlodipine monotherapy arms in the NCT00558428 study Using the standard error in the NCT00558428 study and the standard errors from the regression model in the final model derived in Step 1 permitted derivation of 95% confidence intervals for the treatment effects of interest

Table 3. Blood pressure outcomes based on the simulated treatment comparison.

	Predicted in the NCT00558428 population based on the models fitted in the BA5-A5 trial	Observed in the NCT00558428 study	Estimated via the simulated treatment comparison
Baseline, mmHg (SE)			Not applicable
Systolic BP
Treatment^a	149.3 (0.7)	149.3 (0.7)
Amlodipine 5 mg	150.5 (0.8)	150.5 (0.8)
Diastolic BP
Treatment^a	96.5 (0.3)	96.5 (0.3)
Amlodipine 5 mg	96.4 (0.3)	96.4 (0.3)
At week 8, mmHg (SE)			Not applicable
Systolic BP
Treatment^a	133.6 (1.2)	138.2 (NA)
Amlodipine 5 mg	141.3 (1.3)	144.0 (NA)
Diastolic BP
Treatment^a	82.8 (0.9)	88.6 (NA)
Amlodipine 5 mg	88.3 (0.9)	90.7 (NA)
Difference in BP decrease from baseline to 8 weeks, mmHg (SE) ^b
	Combination	Amlodipine 10 mg	Combination
	versus	versus	versus
	amlodipine 5 mg	amlodipine 5 mg	amlodipine 10 mg
Systolic BP	–6.5 (1.8)	–4.9 (1.0)	–1.6 (1.9)
Diastolic BP	–5.5 (1.2)	–2.2 (0.7)	–3.3 (1.3)

^aTreatment means the combination for the column “predicted”, and amlodipine 10 mg for the column “observed”.

^bThe final model for SBP and DBP included the following variables: baseline SBP, baseline DBP, duration of hypertension, age, concomitant diabetes, sex, smoking history (final model for SBP only), and body mass index (final model for DBP only).

Abbreviations. BP, blood pressure; NA, not available; SE, standard error.

Figure 2. Differences in BP decrease from baseline to 8 weeks for amlodipine 5 mg/bisoprolol 5 mg combination therapy versus amlodipine 10 mg monotherapy.

Abbreviations. BP, Blood pressure; CI, Confidence intervals; SE, standard error. “Main analysis”: the final model for SBP and DBP included the following variables: baseline SBP, baseline DBP, duration of hypertension, age, concomitant diabetes, sex, smoking history (final model for SBP only), and body mass index (final model for DBP only). “Sensitivity: without adjustment” refers to a standard indirect comparison, without any population adjustment. “Sensitivity: all variables” refers to an analysis which includes all selected prognostic variables, without variable selection meaning with addition of body mass index for SBP and smoking history for DBP (Step 1 in Table 2).