confidential: for review only · confidential: for review only the incremental effects of...

Confidential: For Review Only

The incremental effects of antihypertensive drugs: an

instrumental variable analysis

Journal: BMJ

Manuscript ID BMJ.2017.038957.R2

Article Type: Research

BMJ Journal: BMJ

Date Submitted by the Author: 20-Sep-2017

Complete List of Authors: Markovitz, Adam; University of Michigan Medical School; University of Michigan School of Public Health, Health Management and Policy Mack, Jacob; University of Michigan Medical School Nallamothu, Brahmajee; Ann Arbor Health Services Research and Development Center of Excellence, Division of Cardiovascular Medicine;

University of Michigan Medical School, Center for Health Outcomes and Policy Ayanian, John; University of Michigan School of Medicine, Internal Medicine; Institute for Healthcare Policy and Innovation Ryan, Andrew; University of Michigan, Department of Health Management and Policy, School of Public Health

Keywords: Hypertension, Blood Pressure, Antihypertensive Agents, Cardiovascular Diseases, Quasi-Experimental Design

https://mc.manuscriptcentral.com/bmj

BMJ


The incremental effects of antihypertensive drugs: an instrumental variable

analysis

Adam A Markovitz, Jacob A Mack, Brahmajee K Nallamothu, John Z Ayanian, Andrew

M Ryan

University of Michigan Medical School and School of Public Health, 1415 Washington

Heights, Ann Arbor, MI, US

Adam Markovitz

MD/PhD candidate

University of Michigan Medical School, 1301 Catherine St., Ann Arbor, MI, US

Jacob A Mack

junior doctor

University of Michigan Medical School, Ann Arbor, MI, US

Brahmajee K Nallamothu

professor

University of Michigan Institute for Health Policy and Innovation, 2800 Plymouth Rd,

Ann Arbor, MI, US

John Z Ayanian

professor

University of Michigan School of Public Health

Andrew M Ryan

associate professor

Correspondence to:

Andrew M Ryan

Department of Health Management & Policy

University of Michigan School of Public Health

Address: M3124 SPH II, 1415 Washington Heights, Ann Arbor, MI 48109

[email protected]

Phone: 734-936-1311

Fax: 734-764-4338

Page 1 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


What this paper adds

What is already known on this topic.

• Patients with hypertension frequently require multiple antihypertensive drugs.

• Observational studies suggest that benefits diminish when prescribing additional

antihypertensive drugs.

• Yet these results may be due to confounding by indication—when treatments appear

less effective if used in sicker patients.

What this study adds.

• This study suggests that adding antihypertensive drugs results in large reductions in

blood pressure that are similar in magnitude when adding a first, second, third, or fourth

or more drug class to a patient’s regimen.

• These results challenge the view that the effects of antihypertensive drugs will diminish

with each added drug.

Page 2 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Structured Abstract

Objectives. To assess the incremental effects of adding antihypertensive drugs to a

patient’s regimen while accounting for confounding by indication – when treatments

appear less effective if used in sicker patients.

Design. Instrumental variable analysis of Systolic Blood Pressure Intervention Trial

(SPRINT) data. To account for confounding, we used randomization status as the

instrumental variable, with random assignment to intensive (systolic blood pressure

target < 120 mm Hg) versus standard (systolic blood pressure target < 140 mm Hg)

treatment independently increasing the probability of a new drug class being added.

Setting. Secondary data analysis of a randomized clinical trial conducted at 102 sites

between 2010 and 2015.

Participants. 9,092 SPRINT participants with hypertension and increased

cardiovascular risk but no history of diabetes or stroke.

Main outcomes measures. Systolic blood pressure, major cardiovascular events, and

serious adverse events.

Results. In standard multivariable models not adjusted for confounding by indication,

adding antihypertensive drugs was associated with modestly lower systolic blood

Page 3 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


pressure (-1.3 mm Hg, 95% confidence interval [CI], -1.6 to -1.0) and no change in

major cardiovascular events (absolute risk [AR] events per 1000 patient-years, 0.5, 95%

CI, -1.5 to 2.3). In instrumental variable models, antihypertensive drugs led to clinically

important reductions in systolic blood pressure (-14.4 mm Hg, 95% CI, -15.6 to -13.3)

and fewer major cardiovascular events (AR, -6.2, 95% CI, -10.9 to -1.3). Incremental

reductions in systolic blood pressure remained large and similar in magnitude when

adding a first, second, third, or fourth or more antihypertensive drug class to a patient’s

regimen. This finding was consistent across all patient subgroups. Adding

antihypertensive drugs was not associated with adverse events in either standard or

instrumental variable models.

Conclusions. After accounting for confounding by indication, adding antihypertensive

drugs led to large reductions in systolic blood pressure and major cardiovascular events

among patients at high risk for cardiovascular events but without diabetes. Systolic

blood pressure effects persisted across all levels of baseline drug use and all patient

subgroups.

Page 4 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Introduction

Hypertension is a pervasive and growing threat to global health. The number of adults

with hypertension has nearly doubled from 442 million to 874 million worldwide in the

past 25 years.1 Among those with hypertension, many take multiple antihypertensive

drugs to control their blood pressure and reduce cardiovascular risk.2 Although the

addition of a new antihypertensive drug class is done routinely in clinical practice, the

incremental benefits and risks of each additional drug class remain controversial. It is

commonly believed that the addition of a new drug to a patient’s regimen will result in

progressively smaller reductions in blood pressure while increasing the risk of adverse

events.3–5 Due to physiologic limitations or drug-drug interactions, escalating the

number of antihypertensive drugs has been postulated to produce diminishing benefits

and increasing harms.6 Such concerns are particularly pronounced for older patients7,8

or those for whom some drugs may prove less effective, such as patients with resistant

hypertension 9–11 or black patients.12,13 However, others have argued adding a new drug

class that targets a distinct and complementary mechanism may reduce blood pressure

at lower doses of each drug, improving therapeutic benefit and lowering side effects.14,15

Evidence to support either of these ideas is limited. Observational studies suggest that

incremental improvements to hypertension control diminish across successively greater

numbers of drugs2,16,17 and that, beyond a certain point, such gains may be

accompanied by increased risk of adverse events (e.g., injurious fall) and

cardiovascular events (e.g., myocardial infarction).8,9,18,19 Yet these results may be

Page 5 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


confounded by indication. This occurs when a treatment appears harmful or less

effective because it is used in sicker patients. Meta-analyses of randomized trials have

relied largely on studies comparing monotherapy to dual combination therapy, drawing

mixed conclusions regarding whether adding a second drug class produces additive15,20

or diminishing21 incremental effects on systolic blood pressure. As a result, there is a

lack of data on the incremental effects of successively higher numbers of

antihypertensive drugs, particularly when considering the use of three or more drugs in

a patient’s regimen.5

The Systolic Blood Pressure Intervention Trial (SPRINT) presents a unique opportunity

to study this question.22 The purpose of the trial was to compare intensive blood

pressure treatment (i.e., systolic blood pressure target of less than 120 mm Hg) versus

standard treatment (i.e., systolic blood pressure target of less than 140 mm Hg). The

trial was terminated early after finding that intensive therapy lowered substantially the

risk of major cardiovascular events and mortality. Nonetheless, concerns have been

raised regarding an increased risk of some serious adverse events.23,24 Understanding

how these benefits and risks vary with an increasing number of antihypertensive drugs

increased is critical to guide clinical practice. We conducted a secondary analysis of

SPRINT data to evaluate the incremental effects of antihypertensive drugs. To account

for confounding by indication, we performed an instrumental variable analysis, with

random assignment to intensive versus standard treatment independently increasing

the probability of a new drug class being added to a patient’s regimen.

Page 6 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Methods

Study data. We performed a secondary analysis of SPRINT data. This trial was funded

by the National Institute of Health. SPRINT data are held by the National Heart, Lung,

and Blood Institute (NHLBI) in a formal data repository intended to facilitate the sharing

of data from clinical trials and observational studies (accessed here:

https://biolincc.nhlbi.nih.gov/home/).25 Our research team had no relationship to the

original trial. Data were made available to our research team through our participation in

The New England Journal of Medicine’s “SPRINT Data Analysis Challenge,” a research

competition intended to explore the benefits and challenges of sharing data from clinical

trials.26,27 Data were made available on November 1, 2016, one year earlier than

required by the NHLBI, although participants were still required to receive Institutional

Review Board approval (or exception), and sign a data use agreement. Our study was

deemed exempt from review by the Institutional Research Board of the University of

Michigan. SPRINT data are now available on the NHLBI repository to the public. Prior to

analysing SPRINT data, we created a study plan describing our overall study objectives

and analytic strategy. We provide this study plan, as well as revisions we undertook

upon accessing and analysing SPRINT data, in the online appendix.

Study population. SPRINT included participants at least 50 years of age with a systolic

blood pressure of 130 to 180 mm Hg, no history of diabetes or stroke, and at least one

cardiovascular risk factor (clinical or subclinical cardiovascular disease other than

stroke, chronic kidney disease), a 10-year risk of cardiovascular disease of 15% or

Page 7 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


greater based on the Framingham risk score, or age 75 years or older. We additionally

excluded any patients lacking at least one post-baseline measure of the number of

antihypertensive drug classes. We summarize these inclusion and exclusion criteria in

Online Appendix Table 1, as adapted from the original SPRINT analysis.22

Outcomes. We followed SPRINT’s protocol in selecting and defining our three primary

study outcomes: systolic blood pressure; composite major cardiovascular events; and

composite serious adverse events.22 Systolic blood pressure was defined as each

patient’s final recorded measurement. Major cardiovascular events were defined as a

composite including myocardial infarction, acute coronary syndrome not resulting in

myocardial infarction, stroke, acute decompensated heart failure, or death from

cardiovascular causes. Serious adverse events were defined as a composite including

emergency department evaluations for hypotension, syncope, bradycardia, electrolyte

imbalance, injurious fall, or hospitalizations for acute kidney injury or acute renal failure.

Secondary outcomes included the individual components of composite major

cardiovascular events, the individual components of composite serious adverse events,

and diastolic blood pressure.

Exposure. Our exposure was the recorded number of antihypertensive drug classes

prescribed at the final SPRINT visit. To account for confounding by indication, we

performed an instrumental variable analysis to assess the incremental effects of

antihypertensive drugs. We leveraged SPRINT’s study design, where a participant’s

random assignment to the intensive versus standard treatment arm independently

Page 8 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


increased the probability of a new drug class being added. We then assessed if

incremental effects varied with each added drug by performing both multivariable

adjusted and instrumental variable analyses stratified by the number of antihypertensive

drugs at baseline.

We performed an instrumental variable analysis of SPRINT instead of an intention-to-

treat [ITT] analysis because performing an ITT would not have allowed us to estimate

the incremental effect of antihypertensive drugs. This is because the “treatment” to

which SPRINT randomized patients was not antihypertensive drugs per se – it was a

more intensive systolic blood pressure target (< 120 mm Hg). Thus, a stratified ITT

answers an important but different question – does the effect of randomization to an

intensive vs. standard blood pressure goal vary across baseline number of drugs?

Conversely, by performing an instrumental variable analysis, we instead used

randomization as an instrument to answer a related but clinically distinct question - does

the effect of adding an additional drug vary across baseline number of drugs?

Statistical analysis. We first analysed associations between the number of

antihypertensive drug classes and outcomes using standard multivariable adjusted

models. To account for observed differences across baseline antihypertensive drug use,

we adjusted for a rich set of baseline characteristics including age, sex, race, body

mass index (BMI), smoking, high-density lipoprotein (HDL), serum creatinine, statin

lipid-lowering drug use, history of cardiovascular disease, and history of chronic kidney

disease. We estimated ordinary least squares regression models for blood pressure

Page 9 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


analyses. We estimated Aalen additive hazards models for cardiovascular and adverse

event analyses to account for the right-censored nature of survival outcome.28

We then performed instrumental variable models to address confounding by indication.

We estimated two-stage least squares models for our blood pressure analyses,

estimating blood pressure as a function of predicted number of antihypertensive drug

classes (due to randomization status) and covariates (described above). We specified

robust standard errors with clustering at the trial site level. For our cardiovascular and

adverse event models, we estimated a recently validated two-stage additive hazards

model.29 We based statistical inferences on 2,000 nonparametric bootstrap samples to

account for the combined statistical uncertainty of the first- and second-stage

regressions (see Online Appendix Methods for further details).

Next, we assessed if the incremental effects on blood pressure varied when adding a

first, second, third, or fourth or more antihypertensive drug to a patient’s regimen. We

conducted stratified analyses according to the baseline number of antihypertensive drug

classes and then compared whether these incremental effects differed across baseline

strata (see Online Appendix Methods for further details).

Finally, to confirm that our instrument (randomization status) met the three conditions

for validity, we conducted sensitivity analyses to verify that our instrument: (1) was

highly correlated with the exposure (number of antihypertensive drug classes); (2) was

random (no confounding factor jointly affecting the instrument and outcome); and (3) did

Page 10 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


not affect the outcome independent of the exposure. Because SPRINT patients and

physicians were un-blinded to randomization status (although study outcomes were

adjudicated by blinded committee), we examined whether randomization led to

differential exposure to other interventions (e.g., smoking cessation, dietary changes,

more frequent office visits and blood pressure measurements) whose effects would be

subsequently attributed to antihypertensive drugs. Specifically, we evaluated:

incremental effects among patients who were unlikely to receive behavioural

interventions (i.e., not obese or never-smokers at baseline); incremental effects at times

we considered too early in the study period to be likely driven by behavioural

interventions (i.e., at the three-month visit); and incremental effects on total number of

blood pressure measurements over the study period. We also tested the additive

hazards model assumption that effects are additive and constant over time via

nonparametric goodness-of-fit tests and plots of observed cumulative effect estimates

(see online appendix Methods for details).

Sensitivity analyses. In accordance with SPRINT’s protocol,22 we evaluated variation

in the incremental effects of antihypertensive drugs by estimating models stratified by

the following patient characteristics: age; sex; race; obesity; smoking; history of

cardiovascular disease; and history of chronic kidney disease. We also evaluated

incremental effects of adding a fifth or more antihypertensive drug class to a patient’s

regimen. Because SPRINT’s randomization should ensure that the instrument was not

confounded by patient covariates, we estimated instrumental variable models without

further adjustment for patient covariates. Finally, we estimated reduced-form models of

Page 11 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


the effect of randomization status on the number of drug classes prescribed and

associated changes in systolic blood pressure.

Statistical analyses were performed using Stata version 14.1 and the R timereg and

survival packages.30

Patient involvement. No patients were involved in setting the research question or the

outcome measures, nor were they involved in developing plans for recruitment, design,

or implementation of the study. No patients were asked to advise on interpretation or

writing up of results. There are no specific plans to disseminate the results of the

research to study participants or the relevant patient community beyond the usual

channels of publication.

Results

Patient characteristics and antihypertensive drug use. We identified 9,092 SPRINT

patients who met our inclusion criteria (CONSORT diagram in online appendix figure 1).

Median follow-up was 3.25 years. At baseline, 861 (9%) patients were taking no

antihypertensive drug classes, 2,662 (29%) were taking one, 3,201 (35%) were taking

two, and 2,368 (26%) were taking three or more (table 1). All patient characteristics

except smoking status varied across categories of baseline drug use. By the study’s

end, 855 (19%) and 242 (5%) of patients in the intensive treatment group were taking

four and five drug classes, respectively (online appendix table 2).

Page 12 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Instrument validity. We first confirmed that our instrumental variable analysis was

necessary: standard multivariable adjusted models yielded biased estimates relative to

those from instrumental variable models via the C or difference-in-Sargan test of

endogeneity (F1,101 statistic = 1,547; P<0.001).31 We next verified that our instrument

met the three conditions for validity. First, randomization status was highly correlated

with number of antihypertensive drug classes (F1,101=1,065, P<0.001), where

instruments with F statistics above 10 are considered strong (online appendix figure 2

and tables 3 and 4).32 Second, SPRINT’s randomization of intensive versus standard

treatment was successful, with patient characteristics balanced across the two groups

(online appendix table 5). Third, we verified that randomization did not affect patient

outcomes independent of the antihypertensive drug classes by confirming: (1) additive

incremental effects held true for patient populations regardless of their propensity to

receive behavioural interventions, i.e., obese versus non-obese, history of smoking

versus no history of smoking (online appendix figures 3E and 3F); (2) similarly additive

effects occurred at times too early to be driven largely by behavioural interventions, i.e.,

the study’s three-month visit versus averaged over the entire study period (online

appendix table 6 and figures 4 and 5); and (3) incremental changes in antihypertensive

drug use did not affect the total number of blood pressure measurements, which could

otherwise induce a spurious relationship between changes in drug use and blood

pressure (online appendix table 7). Collectively, these results suggest that the

differential effects of randomization status operated through changes in the use of

antihypertensive drugs and not through non-pharmacologic changes. Additionally, we

Page 13 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


confirmed that the incremental effects of antihypertensive drugs and other covariates

were constant and additive over time (online appendix table 8 and figure 6).

Systolic blood pressure. Figure 1 shows the incremental effects of antihypertensive

drugs on systolic blood pressure. In standard multivariable models, adding

antihypertensive drugs was associated with modestly lowered systolic blood pressure (-

1.3 mm Hg, 95% confidence intervals [CI]: -1.6 to -1.0). These effects diminished with

each additional drug (interaction term P<0.001 online appendix table 9). In instrumental

variable models, we estimated a much larger effect of antihypertensive drugs on systolic

blood pressure reductions (-14.4 mm Hg, 95% CI, -15.6 to -13.3). These effects

remained large and similar in magnitude when adding a first drug (-13.9 mm Hg, 95%

CI, -16.3 to -11.5); second drug (-14.2 mm Hg, 95% CI, -16.0 to -12.4); third drug (-14.7

mm Hg, 95% CI, -16.4 to 13.1); or fourth drug or more (-15.1 mm Hg, 95% CI, -17.6 to -

12.6; figure 1 and online appendix table 9). A similar pattern was observed for diastolic

blood pressure: the addition of antihypertensive drugs was associated with small,

diminishing effects in standard multivariable models and with large, approximately

additive effects in instrumental variable models (online appendix table 9 and figure 7).

Major cardiovascular events. Figure 2 shows the incremental effects of

antihypertensive drugs on major cardiovascular events. In standard multivariable

models, antihypertensive drugs were not associated with differences in the risk of

composite cardiovascular events (absolute risk [AR], 0.5 events per 1,000 patient-years;

CI, -1.5 to 2.3) or any component outcomes (online appendix figure 8). Conversely, our

Page 14 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


instrumental variable models suggested that antihypertensive drugs caused a large

reduction in composite major cardiovascular events (AR, -6.2, 95% CI, -10.9 to -1.3)

(figure 2). These effects did not vary systematically across levels of baseline drug use.

In these models, adding a drug also reduced risk of heart failure (AR, -4.0, 95% CI, -6.6

to -1.5), death from cardiovascular causes (AR, -1.8, 95% CI, -3.6 to -0.0), and death

from any cause or a composite cardiovascular outcome (AR, -7.0, 95% CI, -12.4 to -1.6)

(online appendix figure 8).

Serious adverse events. The addition of antihypertensive drugs was not associated

with increased risk of composite serious adverse events in either standard multivariable

models (AR, 4.9, 95% CI, -2.1 to 12.1) or instrumental variable models (AR, 12.7, 95%

CI, -5.0 to 31.0; figure 3). However, our instrumental variable models suggested that

adding antihypertensive drugs increased the risk of three component adverse events:

hypotension (AR, 4.3, 95% CI, 1.4 to 7.3); electrolyte imbalance (AR, 5.5, 95% CI, 2.2

to 8.9); and acute kidney injury or renal failure (AR, 8.3, 95% CI, 4.4 to 12.3; online

appendix figure 9).

Sensitivity analyses. Table 2 shows the incremental effects of antihypertensive drugs

stratified by pre-specified clinical and demographic characteristics. We did not observe

consistent heterogeneity in the effects on blood pressure, cardiovascular events, or

adverse events. In analyses stratified by both baseline drug use and patient

characteristics, we observed consistent, additive reductions in systolic blood pressure

within each patient subgroup, as defined by age, sex, race, obesity, smoking, history of

Page 15 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


cardiovascular disease, or history of chronic kidney disease (online appendix table 10

and figures 3A-G). We also found that incremental effects were robust across our four

alternative specifications: when the exposure was defined by the number of drug

classes at the three-month visit (online appendix table 6 and figure 4); when the

exposure was defined by the mean number of drug classes over the study period

(online appendix table 6 and figure 5); when no adjustment for patient covariates was

made (online appendix figure 10); and when a fifth or more drug class was added to a

patient’s regimen (online appendix figure 11). In reduced-form analyses, we found that

randomization to the intensive group resulted in greater changes to both

antihypertensive drug use and blood pressure among patients who were on relatively

fewer drug classes at baseline (online appendix table 11). This may reflect the fact that

patients on fewer drug classes at baseline had higher baseline blood pressure (table 1)

and thus necessitated greater treatment intensification (and blood pressure reductions)

to reach either the standard or intensive blood pressure targets.

Discussion

Principal findings. Our instrumental variable analysis of SPRINT data found that

adding antihypertensive drugs led to clinically important reductions in systolic blood

pressure and risk of major cardiovascular events but no differences in serious adverse

events. Incremental reductions in systolic blood pressure remained large and similar in

magnitude when adding a first, second, third, or fourth or more antihypertensive drug

class to a patient’s regimen. This finding was consistent for each patient subpopulation

Page 16 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


tested. Collectively, these findings suggest that the incremental effects of

antihypertensive drugs persist across clinically important subgroups and up to four or

more drug classes. Our findings provide physicians, patients, and policymakers with

more rigorous and nuanced insights regarding optimal approaches to managing

hypertension and cardiovascular risk.

Comparison with other studies. If physicians believe that the addition of new

antihypertensive drug classes will result in progressively smaller reductions in blood

pressure,3–6,33 they may be more cautious about adding antihypertensive drugs.

However, investigators have proposed two competing models: additive effects, in which

the combined effect of antihypertensive drugs is equal to the sum of the drugs’

individual effects;15 and synergistic effects, in which the rational addition of a drug

targeting a new, complementary mechanism yields effects greater than the sum of the

drugs’ individual effects.34 Meta-analyses of clinical trials comparing dual versus

monotherapy have found either additive15,20 or diminishing effects,21 while safety and

efficacy trials of some specific triple combination therapies have found short-term

effects that are less than additive.9,35–37

Our study clarifies and extends these prior analyses, leveraging SPRINT’s random

assignment to evaluate longer-run effects across a broad range of antihypertensive

drug classes. We found large reductions in blood pressure that do not decrease as each

additional antihypertensive is added. Further, these additive effects appear robust

across a wide range of patient characteristics. Our findings are particularly relevant

Page 17 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


given longstanding concerns regarding those who may have fewer effective treatment

options, including black patients (e.g., due to decreased responses to ACE inhibitors)

and older patients (e.g., due to increased resistant hypertension).10,38,39 The large,

additive reductions in systolic blood pressure we observed in patients greater than 75

years of age may also help to explain the strong mortality benefits previously reported in

older SPRINT patients.40 This last observation is particularly important given the rising

share of adults now using at least three classes of antihypertensive drugs.41

There is a paucity of direct evidence on the incremental effects of antihypertensive

drugs on risk of cardiovascular events and serious adverse events.15 Short-term (8-12

week) safety and efficacy trials of triple combination therapy, meanwhile, have found

diminishing short-term effects on risk of adverse events.9,35–37 Our study adds new

evidence on the longer-run effects of antihypertensive drugs with the median follow-up

of 3.25 years in SPRINT. Similar to the original SPRINT study, we found that the

addition of an antihypertensive reduces the risk of composite major cardiovascular

events (driven by decreased risk of heart failure) while increasing the risk of

hypotension, electrolyte imbalance, and acute kidney injury or renal failure. In stratified

analyses, we found that the incremental effects on cardiovascular and adverse event

risk, though estimated with less precision, do not systematically differ across levels of

baseline drug use.

Strengths and limitations of study. The key strength of this study is its use of

randomization from a clinical trial to examine a question critical to clinical practice but

Page 18 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


typically confounded by indication: what are the incremental effects of adding

antihypertensive drugs to a patient’s current regimen? By performing an instrumental

variable analysis in the context of a clinical trial, our study harnessed a particularly

strong instrument – randomization status itself – to assess the effects of otherwise non-

random changes in prescribing behavior. SPRINT’s study design also allowed us to

examine these incremental effects across a wider range of baseline drug use, a more

diverse set of drug classes, and over a longer period of follow-up than is typically

assessed in trials of combination drug therapy. In addition, the excellent data capture in

the SPRINT trial limits bias from non-random samples attrition.

In comparing results from our instrumental variable models to those from standard

multivariable regression models, we shed light on a pervasive source of bias in

observational medical studies – confounding by indication. This confounding

dramatically changed estimates in our study, positively biasing our estimates to such an

extent that antihypertensive drugs appear to have almost no effect on blood pressure

(from -1.3 mm Hg in standard models to -14.4 mm Hg in instrumental variable models)

and in fact increase the risk of major cardiovascular events (from +0.5 to -6.2 events per

1,000 patient-years). Importantly, this bias persists after adjusting for a rich set of

baseline clinical covariates that are not always available to researchers, including BMI,

smoking status, HDL, serum creatinine, statin lipid-lowering drug use, and history of

cardiovascular and chronic kidney disease. The nature of confounding by indication in

our study predicts the direction of this bias: because the omitted variable (more severe

hypertension, elevated cardiovascular risk) is positively related to both the exposure

Page 19 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


(number of antihypertensive drugs) and the outcome (blood pressure, cardiovascular

event incidence), failing to account for the confounder will positively bias estimates of

the true effect (described by the “omitted variable bias formula”42). Conversely, in

instrumental variable models that are robust to confounding by indication, we find that

the addition of antihypertensive drugs leads to large decreases in blood pressure and

risk of cardiovascular events.

Several important limitations to our study also merit discussion. First, because we

observed only the number of prescribed drug classes and not drug dosage or

adherence, we cannot exclude the possibility that randomization led to differential

changes in drug dosage or adherence whose effects would be subsequently attributed

to changes in the number of drug classes. However, it is unclear whether randomization

to the intensive group and subsequent addition of a new drug class would be

accompanied by the clinician: (1) increasing the dose of the first drug to lower blood

pressure further (negatively biasing estimates away from zero by attributing this dose

change to the addition of a new drug class); or (2) lowering the dose of either the first or

added drug class to take advantage of complementary mechanisms and widening

therapeutic windows (positively biasing estimates toward zero). Thus, the net effect in

our study remains uncertain. These data limitations also precluded our ability to analyse

whether certain drug combinations or intensification sequences were more successful

than others. Nonetheless, supplemental analysis suggests that the blood pressure

effects of antihypertensive drugs do not attenuate across patients receiving five or more

drug classes (online appendix figure 11). Moreover, the original SPRINT analysis

Page 20 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


reported a diverse range of prescribed drug classes: along with first-line agents such as

thiazide-type diuretics (67% of patients by the study’s end), calcium channel blockers

(57%), and angiotensin-converting-enzyme inhibitors/angiotensin II receptor blockers

(77%), SPRINT patients were also prescribed beta-blockers (41%), aldosterone

receptor blocker diuretics (9%), and direct vasodilators (7%), among others.22 Given

that 19% of patients in the intensive group were on four drug classes and 5% were on

five drug classes at the study’s conclusion, these data suggest that the additive effects

we observed were not entirely due to first-line agents.

Second, we cannot ascertain whether patients in the intensive group were differentially

exposed to behavioural interventions such as dietary changes or smoking cessation).

Nonetheless, sensitivity analyses demonstrated similar effects among those patients

who were unlikely to receive such behavioural interventions (i.e., non-obese and non-

smoking at baseline) and at times too early to be likely driven by nonpharmacologic

means (i.e., three-month visit). Third, our results may not generalize to several high-risk

populations excluded from SPRINT, including those with diabetes or prior stroke.43,44

Similarly, because SPRINT patients were screened for nonadherence, incremental

effects may be smaller in real-world settings where adherence is lower.5,45 While our

systolic blood pressure estimate is larger than those previously reported for a “standard

dose” (reductions of 14.8 versus 9.1 mm Hg20, respectively), our results parallel

SPRINT’s finding that the intensive treatment arm achieved a 14.8 mm Hg greater

reduction than the standard treatment arm and that doing so required, on average, one

more drug class.22 Finally, our analyses of cardiovascular and adverse events may have

Page 21 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


had insufficient power to detect variation in incremental effects across levels of baseline

drug use. However, we see no clear pattern of evidence suggestive of effect

modification by baseline drug use.

Conclusions and policy implications. These limitations notwithstanding, our findings

provide important new evidence on the incremental effects of antihypertensive drugs.

Our results challenge the view that adding antihypertensive drugs will result in

progressively diminishing effects on blood pressure and cardiovascular events. Our

findings provide patients and clinicians with more rigorous, nuanced insight into optimal

management of hypertension. For clinical and health services researchers, we

demonstrate the importance of accounting for confounding by indication and the value

of using randomization from clinical trials to understand important causal pathways. For

policymakers, our results inform efforts to model and improve cost-effective treatment

for hypertension, a condition with annual costs exceeding $370 billion globally.46

Evidence on the incremental effects of combinations of three or more antihypertensive

drugs is particularly critical for evaluating the cost-effectiveness of widespread treat-to-

target recommendations and the polypharmacy that such policies may require.5

Collectively, these findings suggest that antihypertensive drugs can be used to lower

blood pressure effectively when adding a first, second, third, or fourth or more drug

class to a patient’s regimen. Future research in this area should focus on developing

experimental evidence on how the strong, additive reductions in blood pressure

observed in our study may translate to patient benefit and harm.

Page 22 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Acknowledgements. We wish to thank Dr. J. Brian Byrd and the BMJ reviewers for the

thoughtful guidance and critical review of the manuscript. We would also like to thank

the SPRINT research team for making trial data available to external researchers in a

timely and transparent manner.

Contributors. AAM had full access to all of the study data and takes responsibility for

the integrity of the data and the accuracy of the data analysis. He is the guarantor.

AAM, JAM, BKN, and AMR conceived and designed the study, JAM and BKN acquired

the data, AAM performed the analysis, and all authors provided intellectual input,

analysed and interpreted the data, critically revised the manuscript, and gave final

approval of the manuscript.

Competing interests. We have read and understood BMJ policy on declaration of

interests and declare: AMR has received research grants from the NIH (R01 AG047932

02) for the submitted work; Dr. Nallamothu receives support from the American Heart

Association for his role as Editor-in-Chief of Circulation: Cardiovascular Quality &

Outcomes and has previously served on the Scientific Cardiac Advisory Board for

United Healthcare. No funding organization was involved in the design and conduct of

the study; collection, management, analysis, and interpretation of the data; preparation,

review, or approval of the manuscript; and decision to submit the manuscript for

publication

Page 23 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Data sharing. SPRINT data are available on the NHLBI repository (accessed here:

https://biolincc.nhlbi.nih.gov/home/).

Transparency. The lead author affirms that the manuscript is an honest, accurate, and

transparent account of the study being reported; that no important aspects of the study

have been omitted; and that any discrepancies from the study as planned (and, if

relevant, registered) have been explained.

Exclusive licence. The Corresponding Author has the right to grant on behalf of all

authors and does grant on behalf of all authors, a worldwide licence to the Publishers

and its licensees in perpetuity, in all forms, formats and media (whether known now or

created in the future), to i) publish, reproduce, distribute, display and store the

Contribution, ii) translate the Contribution into other languages, create adaptations,

reprints, include within collections and create summaries, extracts and/or, abstracts of

the Contribution, iii) create any other derivative work(s) based on the Contribution, iv) to

exploit all subsidiary rights in the Contribution, v) the inclusion of electronic links from

the Contribution to third party material where-ever it may be located; and, vi) licence any

third party to do any or all of the above.

Page 24 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


References

1. Forouzanfar MH, Liu P, Roth GA, et al. Global Burden of Hypertension and Systolic Blood Pressure of at Least 110 to 115 mm Hg, 1990-2015. JAMA. 2017;317(2):827-838. doi:10.1001/JAMA.2016.19043.

2. Gu Q, Burt VL, Dillon CF, Yoon S. Trends in antihypertensive medication use and blood pressure control among United States adults with hypertension: the National Health And Nutrition Examination Survey, 2001 to 2010. Circulation. 2012;126(17):2105-2114. http://circ.ahajournals.org/cgi/doi/10.1161/CIRCULATIONAHA.112.096156.

3. Rana BK, Dhamija A, Panizzon MS, et al. Imputing observed blood pressure for antihypertensive treatment: impact on population and genetic analyses. Am J Hypertens. 2014;27(6):828-837. doi:10.1093/ajh/hpt271.

4. Timbie JW, Hayward RA, Vijan S. Variation in the net benefit of aggressive cardiovascular risk factor control across the US population of patients with diabetes mellitus. Arch Intern Med. 2010;170(12):1037-1044. doi:10.1001/archinternmed.2010.150.

5. Timbie JW, Hayward RA, Vijan S. Diminishing efficacy of combination therapy, response-heterogeneity, and treatment intolerance limit the attainability of tight risk factor control in patients with diabetes. Heal Serv Res. 2010;45(2):437-456. doi:10.1111/j.1475-6773.2009.01075.x.

6. Mold JW, Hamm RM, McCarthy LH. The Law of Diminishing Returns in Clinical Medicine: How Much Risk Reduction is Enough? J Am Board Fam Med. 2010;23(3):371-375. http://www.jabfm.org/cgi/doi/10.3122/jabfm.2010.03.090178.

7. Maher RL, Hanlon J, Hajjar ER. Clinical consequences of polypharmacy in elderly. Expert Opin Drug Saf. 2013;13(1):57-65. doi:10.1517/14740338.2013.827660.

8. Mukete BN, Ferdinand KC. Polypharmacy in Older Adults With Hypertension: A Comprehensive Review. J Clin Hypertens. 2015;18(1):10-18. http://doi.wiley.com/10.1111/jch.12624.

9. Calhoun DA, Lacourciere Y, Chiang YT, Glazer RD. Triple antihypertensive therapy with amlodipine, valsartan, and hydrochlorothiazide: a randomized clinical trial. Hypertension. 2009;54(1):32-39. doi:10.1161/HYPERTENSIONAHA.109.131300.

10. James PA, Oparil S, Carter BL, et al. 2014 Evidence-Based Guideline for the Management of High Blood Pressure in Adults. JAMA. 2014;311(5):507-514. http://jama.jamanetwork.com/article.aspx?doi=10.1001/jama.2013.284427.

11. Hypertension EETF for the M of A, Mancia G, Fagard R, et al. 2013 Practice guidelines for the management of arterial hypertension of the European Society of Hypertension (ESH) and the European Society of Cardiology (ESC): ESH/ESC Task Force for the Management of Arterial Hypertension. J Hypertens. 2013;31(10):1925-1938. doi:10.1097/HJH.0b013e328364ca4c.

12. Calhoun DA, Jones D, Textor S, et al. Resistant hypertension: diagnosis, evaluation, and treatment: a scientific statement from the American Heart Association Professional Education Committee of the Council for High Blood Pressure Research. Circulation. 2008;117(25):e510-26.

Page 25 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


doi:10.1161/CIRCULATIONAHA.108.189141. 13. Flack JM, Sica DA, Bakris G, et al. Management of high blood pressure in Blacks:

an update of the International Society on Hypertension in Blacks consensus statement. Hypertension. 2010;56(5):780-800. doi:10.1161/HYPERTENSIONAHA.110.152892.

14. Gradman AH, Basile JN, Carter BL, Bakris GL, Group on behalf of the AS of HW. Combination Therapy in Hypertension. J Clin Hypertens. 2010;13(3):146-154. http://doi.wiley.com/10.1111/j.1751-7176.2010.00397.x.

15. Law MR, Wald NJ, Morris JK, Jordan RE. Value of low dose combination treatment with blood pressure lowering drugs: analysis of 354 randomised trials. BMJ. 2003;326(7404):1427. doi:10.1136/bmj.326.7404.1427.

16. Egan BM, Li J, Shatat IF, Fuller JM, Sinopoli A. Closing the Gap in Hypertension Control Between Younger and Older Adults: National Health and Nutrition Examination Survey (NHANES) 1988 to 2010. Circulation. 2014;129(20):2052-2061. http://circ.ahajournals.org/cgi/doi/10.1161/CIRCULATIONAHA.113.007699.

17. Murphy CM, Kearney PM, Shelley EB, Fahey T, Dooley C, Kenny RA. Hypertension prevalence, awareness, treatment and control in the over 50s in Ireland: evidence from The Irish Longitudinal Study on Ageing. J Public Heal. 2016;38(3):450-458. doi:10.1093/pubmed/fdv057.

18. Cruickshank J, Thorp J, Zacharias FJ. Benefits and potential harm of lowering high blood pressure. Lancet. 1987;329:581-584.

19. Zanchetti A. Blood pressure targets of antihypertensive treatment: up and down the J-shaped curve. Eur Hear J. 2010;31(23):2837-2840. doi:10.1093/eurheartj/ehq281.

20. Wald DS, Law M, Morris JK, Bestwick JP, Wald NJ. Combination Therapy Versus Monotherapy in Reducing Blood Pressure: Meta-analysis on 11,000 Participants from 42 Trials. AJM. 2009;122:290-300.

21. Wu J, Kraja AT, Oberman A, et al. A summary of the effects of antihypertensive medications on measured blood pressure. Am J Hypertens. 2005;18(7):935-942. http://ajh.oxfordjournals.org/cgi/doi/10.1016/j.amjhyper.2005.01.011.

22. SPRINT Writing Group, Wright JT, Williamson JD, et al. A Randomized Trial of Intensive versus Standard Blood-Pressure Control. N Engl J Med. 2015;373(22):2103-2116. http://eutils.ncbi.nlm.nih.gov/entrez/eutils/elink.fcgi?dbfrom=pubmed&id=26551272&retmode=ref&cmd=prlinks.

23. Touyz RM, Dominiczak AF. Successes of SPRINT, but Still Some Hurdles to Cross. Hypertension. 2016;67(2):268-269. doi:10.1161/HYPERTENSIONAHA.115.06725.

24. Cushman WC, Whelton PK, Fine LJ, et al. SPRINT Trial Results: Latest News in Hypertension Management. Hypertension. 2016;67(2):263-265. doi:10.1161/HYPERTENSIONAHA.115.06722.

25. Coady SA, Mensah GA, Wagner EL, Goldfarb ME, Hitchcock DM, Giffen CA. Use of the National Heart, Lung, and Blood Institute Data Repository. N Engl J Med. 2017;376(19):1849-1858. doi:10.1056/NEJMsa1603542.

26. Burns NS, Miller PW. Learning What We Didn’t Know — The SPRINT Data Analysis Challenge. N Engl J Med. 2017;363(1):1-3. doi:10.1056/NEJMp1002530.

Page 26 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


27. New England Journal of Medicine. The SPRINT Data Analysis Challenge. https://challenge.nejm.org/pages/home. Published 2017.

28. Aalen OO. A linear regression model for the analysis of life times. Stat Med. 1989;8(8):907-925. http://doi.wiley.com/10.1002/sim.4780080803.

29. Tchetgen Tchetgen EJ, Walter S, Vansteelandt S, Martinussen T, Glymour M. Instrumental Variable Estimation in a Survival Context. Epidemiology. 2015;26(3):402-410. http://content.wkhealth.com/linkback/openurl?sid=WKPTLP:landingpage&an=00001648-201505000-00016.

30. Therneau TM. A package for Survival Analysis in R. 2015. 31. Baum CF, Schaffer ME, Stillman S. Instrumental variables and GMM: Estimation

and testing. Stata J. 2003;3(1):1-31. doi:The Stata Journal. 32. Staiger D, Stock J. Instrumental Variables Regression with Weak Instruments.

Natl Bur Econ Res. 1997;65(3):557-586. 33. Moran AE, Odden MC, Thanataveerat A, et al. Cost-Effectiveness of

Hypertension Therapy According to 2014 Guidelines. N Engl J Med. 2015;372:447-455.

34. G.Y. L, Beevers M, Beevers DG. The “Birmingham Hypertension Square” for the optimum choice of add-in drugs in the management of resistant hypertension. J Hum Hypertens. 1998;12(11):761-763.

35. Ferdinand KC, Weitzman R, Israel M, Lee J, Purkayastha D, Jaimes EA. Efficacy and safety of aliskiren-based dual and triple combination therapies in US minority patients with stage 2 hypertension. J Am Soc Hypertens. 2011;5(2):102-113. http://linkinghub.elsevier.com/retrieve/pii/S1933171111000076.

36. Gradman AH. Rationale for triple-combination therapy for management of high blood pressure. J Clin Hypertens. 2010;12(11):869-878. doi:10.1111/j.1751-7176.2010.00360.x.

37. Oparil S, Melino M, Lee J, Fernandez V, Heyrman R. Triple therapy with olmesartan medoxomil, amlodipine besylate, and hydrochlorothiazide in adult patients with hypertension: The TRINITY multicenter, randomized, double-blind, 12-week, parallel-group study. Clin Ther. 2010;32:1252-1269.

38. Aronow WS, Harrington RA, Fleg JL, et al. ACCF/AHA 2011 Expert Consensus Document on Hypertension in the Elderly. Circulation. 2011;123:2434-2506.

39. Flack JM, Sica DA, Bakris G, et al. Management of high blood pressure in Blacks: an update of the International Society on Hypertension in Blacks consensus statement. Hypertension. 2010;56(5):780-800. doi:10.1161/HYPERTENSIONAHA.110.152892.

40. Williamson JD, Supiano MA, Applegate WB, et al. Intensive vs Standard Blood Pressure Control and Cardiovascular Disease Outcomes in Adults Aged ≥75 Years. JAMA. 2016;315(24):2610-2673. http://jama.jamanetwork.com/article.aspx?doi=10.1001/jama.2016.7050.

41. Bromfield SG, Bowling CB, Tanner RM, et al. Trends in hypertension prevalence, awareness, treatment, and control among US adults 80 years and older, 1988-2010. J Clin Hypertens. 2014;16(4):270-276. doi:10.1111/jch.12281.

42. Clarke KA. The Phantom Menace: Omitted Variable Bias in Econometric Research. Confl Manag Peace Sci. 2005;22(4):341-352.

Page 27 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


doi:10.1080/07388940500339183. 43. Bress AP, Tanner RM, Hess R, Colantonio LD, Shimbo D, Muntner P.

Generalizability of SPRINT Results to the U.S. Adult Population. J Am Coll Cardiol. 2016;67(5):463-472. doi:10.1016/j.jacc.2015.10.037.

44. Navar AM, Pencina MJ, Peterson ED. Assessing Cardiovascular Risk to Guide Hypertension Diagnosis and Treatment. JAMA Cardiol. 2016;27705(8):864-871. doi:10.1001/jamacardio.2016.2861.

45. Nallamothu BK, Hayward RA, Bates ER. Beyond the randomized clinical trial: the role of effectiveness studies in evaluating cardiovascular therapies. Circulation. 2008;118(12):1294-1303. doi:10.1161/CIRCULATIONAHA.107.703579.

46. Gaziano TA, Bitton A, Anand S, Weinstein MC. The global cost of nonoptimal blood pressure. J Hypertens. 2009;27(7):1472-1477. doi:10.1097/HJH.0b013e32832a9ba3.

Page 28 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Table 1. Baseline characteristics of the SPRINT study participants by number of

antihypertensive drugs at baseline.

Overall By number of antihypertensive drug classes at baselinea

Characteristics Total

(N= 9,092)

No drug

classes (N=

861)

One drug

class

(N=2,662)

Two drug

classes (N=

3,201)

Three or

more drug

classes

(N= 2,368)

P value

Age - years (SD) 67.8 (9.4) 65.6 (9.5) 67.6 (9.3) 68.0 (9.5) 68.7 (9.2) <0.001

Female - no. (%) 3,217 (35.4%) 226 (26.2%) 942 (35.4%) 1,151 (36.0%) 898 (37.9%) <0.001

Black race - no. (%)b 2,856 (31.4%) 216 (25.1%) 759 (28.5%) 1,020 (31.9%) 861 (36.4%) <0.001

Body-mass index

(SD)c

29.9 (5.8) 29.0 (5.6) 29.1 (5.5) 30.0 (5.7) 30.9 (6.0) <0.001

Fasting HDL

cholesterol - mg/dl

(SD)

52.8 (14.5) 52.7 (13.9) 53.9 (14.9) 53.0 (14.7) 51.5 (13.7) <0.001

Serum creatinine -

mg/dl (SD)

1.1 (0.3) 1.0 (0.2) 1.0 (0.3) 1.1 (0.3) 1.2 (0.4) <0.001

Statin use - no. (%) 3,970 (43.7%) 175 (20.3%) 1,097 (41.2%) 1,497 (46.8%) 1,201 (50.7%) <0.001

Ever smoker - no. (%) 5,096 (56.0%) 497 (57.7%) 1,459 (54.8%) 1,790 (55.9%) 1,350 (57.0%) 0.310

History of

cardiovascular

disease - no. (%)

1,524 (16.8%) 72 (8.4%) 335 (12.6%) 604 (18.9%) 513 (21.7%) <0.001

Page 29 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


History of chronic

kidney disease - no.

(%)d

2,572 (28.3%) 115 (13.4%) 617 (23.2%) 937 (29.3%) 903 (38.1%) <0.001

Baseline blood

pressure - mm Hg

(SD)

Systolic 139.7 (15.6) 145.1 (15.4) 139.9 (15.3) 138.7 (15.6) 138.8 (15.7) <0.001

Diastolic 78.1 (11.9) 84.2 (11.9) 79.1 (11.2) 77.5 (11.8) 75.7 (12.1) <0.001

SI conversions factor: To convert the values for cholesterol to millimoles per liter, multiply by 0.02586. To convert the

values for glucose to millimoles per liter, multiply by 0.05551. HDL denotes high-density lipoprotein.

aAntihypertensive drugs comprised the following classes: thiazide diuretics; loop diuretics; potassium-sparing diuretics,

aldosterone receptor blockers, beta blockers; beta blockers with intrinsic sympathomimetic activity; combined alpha and

beta blockers; angiotensin-converting-enzyme (ACE) inhibitors; angiotensin II receptor blockers (ARB); non-

dihydropyridine calcium channel blockers; dihydropyridine calcium channel blockers; alpha-1 blockers; central alpha-2

agonists or other centrally acting drugs; direct vasodilators, direct renin inhibitors.

bRace was self-report. Black race includes non-Hispanic and Hispanic black participants.

cBody-mass index is weight in kilograms divided by the square of height in meters.

dNo history of chronic kidney disease includes some participants with unknown chronic kidney disease status at

baseline.

Page 30 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Figure 1. Incremental effects of antihypertensive drugs on systolic blood pressure

Diamonds represent point estimates from pooled models. Squares represent point estimates from

models stratified by baseline number of drug classes. Lines represent 95% confidence intervals.

Systolic blood pressure represents each patient’s final recorded blood pressure measurement.

Antihypertensive drug classes are measured at baseline and at the latest visit for which there was

also a recorded blood pressure measurement. Multivariable adjusted models were estimated using

ordinary least squares regression. Instrumental variable models were estimated using two-stage

ordinary least squares regression.

Page 31 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Figure 2. Incremental effects of antihypertensive drugs on major cardiovascular events



Major cardiovascular events were defined as a composite including myocardial infarction, acute

coronary syndrome not resulting in myocardial infarction, stroke, acute decompensated heart failure,

or death from cardiovascular causes.21 Antihypertensive drug classes are measured at baseline and

at each patient’s final visit. To minimize reverse causality, we used the last recorded value of drug

classes before the incidence of an event for those patients who experienced a major cardiovascular

event. For standard multivariable models, we estimated additive hazards models to account for the

right-censored nature of survival outcomes.26 For instrumental variable models, we implemented a

recently validated two-stage approach,27 substituting the predicted number of drug classes from the

first stage (a function of randomization status and covariates) into an additive hazards model.

Page 32 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Figure 3. Incremental effects of antihypertensive drugs on serious adverse events



Serious adverse events were defined as a composite including emergency department evaluations

for hypotension, syncope, bradycardia, electrolyte imbalance, injurious fall, or hospitalizations for

acute kidney injury or acute renal failure.21 Antihypertensive drug classes are measured at baseline

and at each patient’s final visit. To minimize reverse causality, we used the last recorded value of

drug classes before the incidence of an event for those patient who experienced a serious adverse

event. For standard multivariable models, we estimated additive hazards models to account for the

right-censored nature of survival outcomes.26 For instrumental variable models, we implemented a

recently validated two-stage approach,27 substituting the predicted number of drug classes from the

first stage (a function of randomization status and covariates) into an additive hazards model.

Page 33 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Table 2. Incremental effects of antihypertensive drugs across clinical and demographic subgroups

Subgroup Systolic blood pressure

(mm Hg)

Major cardiovascular

events

(per thousand person-

years)

Serious adverse events

(per thousand person-

years)

N

Incremental effects of antihypertensive drugs (95% confidence interval)

Multivariable Adjusted Models

Overall -1.3 (-1.6, -1) 0.5 (-1.5, 2.3) 4.9 (-2.1, 12.1) 9,092

Age

< 75 yr -1.6 (-1.9, -1.2) 0.5 (-1.2, 2.2) 7.2 (0.2, 14.2) 6,535

≥ 75 yr -0.8 (-1.3, -0.3) 0.7 (-4.7, 5.9) 8.4 (-8.9, 26.2) 2,557

Sex

Female -1.5 (-1.9, -1.1) 0.3 (-2, 2.7) 2.2 (-6.6, 10.7) 5,875

Male -1.1 (-1.6, -0.7) 0.8 (-1.9, 3.7) 15.8 (4.5, 27.4) 3,217

Race

Black -1.5 (-1.8, -1.2) 1.4 (-1, 3.9) 11.5 (3.5, 20.4) 6,236

Nonblack -1 (-1.4, -0.5) -1.4 (-4.2, 1.4) -1.8 (-13.2, 9.2) 2,856

Obesity

Non-obese -1.4 (-1.8, -1.1) 0.4 (-1.9, 2.6) 6.5 (-0.7, 13.6) 7,582

Obese -1.1 (-1.7, -0.5) 1.2 (-2.2, 4.7) 9.1 (-7, 26.4) 1,569

Page 34 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Smoking status

Never smoker -1.1 (-1.5, -0.7) 2.7 (0.1, 5.6) 5 (-4.1, 14.5) 3,996

Ever smoker -1.5 (-1.9, -1.1) -1.3 (-3.9, 1.3) 8.9 (-1, 18.7) 5,096

History of cardiovascular disease

Yes -1.5 (-1.8, -1.1) 0.7 (-1.1, 2.5) 8.9 (1.8, 15.9) 7,568

No -0.7 (-1.4, 0) -0.9 (-8.2, 6.4) -4.1 (-29.6, 22.2) 1,524

History of chronic kidney disease

No -1.7 (-2.1, -1.4) 0.5 (-1.4, 2.5) 4.8 (-2.6, 12.2) 6,520

Yes -0.4 (-0.9, 0.1) -0.3 (-4.5, 4) 11.4 (-4.9, 28) 2,572

Instrumental variable models

Overall -14.4 (-15.6, -13.3) -6.2 (-10.9, -1.3) 12.7 (-5.0, 31.0) 9,092

Age

< 75 yr -14.7 (-16, -13.4) -3.4 (-7.9, 1.1) 16.7 (-0.9, 34.1) 6,535

≥ 75 yr -13.7 (-15.7, -11.7) -15.6 (-30.9, -1.4) 0.1 (-47.5, 44.5) 2,557

Sex

Female -13.7 (-14.9, -12.6) -6.8 (-12.5, -0.9) 4.9 (-15.4, 25.3) 5,875

Male -15.8 (-17.8, -13.8) -4.9 (-13.4, 3.5) 30.3 (-1.5, 61.5) 3,217

Racea

Black -14.6 (-15.9, -13.3) -6.4 (-12.8, -0.2) 19.5 (-2.4, 41.2) 6,236

Page 35 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Nonblack -14.1 (-15.8, -12.4) -5.5 (-13, 2.1) -1.8 (-30.3, 26.5) 2,856

Obesityb

Non-obese -14.5 (-15.6, -13.3) -6.9 (-12.2, -1.6) 12.2 (-7.1, 31.1) 7,582

Obese -13.7 (-16.1, -11.2) -2 (-13.2, 8.5) 14.1 (-27.2, 58.9) 1,569

Smoking status

Never-smoker -1.1 (-1.5, -0.7) -4.9 (-11.5, 1.6) 17.4 (-5.8, 43.3) 3,996

Ever-smoker -1.5 (-1.9, -1.1) -7.3 (-14.1, -0.3) 9.6 (-14.3, 33.4) 5,096

History of cardiovascular disease

Yes -14.5 (-15.8, -13.2) -6.9 (-11.7, -2.5) 11.8 (-5.5, 28.9) 7,568

No -14 (-16.4, -11.5) -0.8 (-21.5, 18.2) 22.2 (-42.2, 89.9) 1,524

History of chronic kidney diseasec

No -14.4 (-15.6, -13.2) -6.9 (-12, -2.1) 17.7 (-0.3, 35.4) 6,520

Yes -14.6 (-16.9, -12.3) -5 (-18.3, 7.9) -6.6 (-48.8, 38.7) 2,572

SI conversions factor: To convert the values for cholesterol to millimoles per liter, multiply by 0.02586. To convert the

values for glucose to millimoles per liter, multiply by 0.05551. HDL denotes high-density lipoprotein.

aRace was self-report. Black race includes Hispanic black and black.

bObesity is defined as a body-mass index greater than or equal to 35.

cNo history of chronic kidney disease includes some participants with unknown chronic kidney disease status at

baseline.

Page 36 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Online Appendix Table of Contents

I. Online Appendix Study Plan II. Online Appendix Table 1. Inclusion and exclusion criteria (adapted from SPRINT analysis

III. Online Appendix Methods IV. Online Appendix Figure 1. CONSORT diagram V. Online Appendix Table 2. Distribution of number of antihypertensive drug classes at each

patient’s final visit VI. Online Appendix Figure 2. Change in number of antihypertensive drug classes over study

VII. Online Appendix Table 3. Change in number of antihypertensive drug classes over study period by number of drug classes at baseline

VIII. Online Appendix Table 4. Instrument strength stratified by number of drug classes at baseline IX. Online Appendix Table 5. Baseline characteristics of the SPRINT study participants by

SPRINT randomization status X. Online Appendix Figure 3. Incremental effects across clinical and demographic subgroups

XI. Online Appendix Table 6. Incremental effects of antihypertensive drugs on systolic blood pressure, major cardiovascular events, and serious adverse events at three-month visit and over the study period

XII. Online Appendix Figure 4. Incremental effect of antihypertensive drugs on systolic blood pressure at the three-month visit

XIII. Online Appendix Figure 5. Incremental effect of mean number of antihypertensive drugs on systolic blood pressure

XIV. Online Appendix Table 7. Incremental effects of antihypertensive drugs on total number of blood pressure measurements

XV. Online Appendix Table 8. Nonparametric tests of constant additive effects on composite major cardiovascular events and serious adverse events

XVI. Online Appendix Figure 6. Cumulative coefficient plots of the incremental effects on composite major cardiovascular events

XVII. Online Appendix Table 9. Differences in incremental effects on blood pressure when adding first, second, third, or fourth or more antihypertensive drug class

XVIII. Online Appendix Figure 7. Incremental effect of antihypertensive drugs on diastolic blood pressure

XIX. Online Appendix Figure 8. Incremental effect of antihypertensive drugs on component major cardiovascular events and all-cause mortality

XX. Online Appendix Figure 9. Incremental effect of antihypertensive drugs on component serious adverse events

XXI. Online Appendix Table 10. Incremental effects on systolic blood pressure across clinical and demographic subgroups

XXII. Online Appendix Figure 10. Incremental effect of antihypertensive drugs on systolic blood pressure in unadjusted models

XXIII. Online Appendix Figure 11. Incremental effect of adding a fifth or more antihypertensive drug class on systolic blood pressure

XXIV. Online Appendix Table 11. Intention-to-treat analyses of the effect of randomization status on changes in prescribed number of drug classes and systolic blood pressure

XXV. Online Appendix References

Page 37 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Online Appendix Study Plan (12/7/16)

Background. Millions of Americans take multiple antihypertensive drugs to control their blood pressure and reduce risk. Although the addition of a new drug is done routinely in clinical practice, its incremental benefits and risks are unknown. It has been presumed that the addition of a new drug typically leads to less improvements in blood pressure but increase the risks of side effects due to potential drug-drug interactions. There is limited empirical evidence to support these ideas. The SPRINT trial presents a unique opportunity to study this question. The purpose of the trial was to

compare intensive therapy for blood pressure with conservative therapy. Although the overall trial

showed a benefit in regards to lower adverse cardiac events and mortality, concerns were raised

given a higher risk of side effects. For a clinical provider it would be very useful to know how these

benefits and risks varied as the number of antihypertensive drugs increased in use.

Accordingly, we will use the design of the SPRINT trial to answer the important question of how the

marginal benefits and risks varied when adding a second, third, or fourth or more drug onto a

patient’s regimen. We will evaluate changes in cardiovascular events, adverse events, and blood

pressure. We will leverage the study design of the SPRINT trial to create an instrument by which

randomization to intensive therapy would increase the likelihood the addition of a new drug and then

measured its marginal effects.

Objective. To assess the incremental effects of adding antihypertensive drugs while accounting for

confounding.

Data. We will perform a secondary data analysis of data collected from the Systolic Blood Pressure

Intervention Trial (SPRINT; n=9361)

Study population. SPRINT inclusion/exclusion criteria:

• Age ≥ 50 • Systolic blood pressure between 130 mm Hg and 180 mm Hg • No history of diabetes or stroke • 1+ cardiovascular risk factor: cardiovascular disease (other than stroke or chronic kidney

disease); FRS ≥ 15%; age ≥ 75.

Participants. 9,092 SPRINT participants with hypertension and increased cardiovascular risk but no

history of diabetes or stroke.

Outcomes. Our main study outcomes are composite major cardiovascular events and composite

serious adverse events. Our secondary outcome is systolic blood pressure. Major cardiovascular

events are defined in SPRINT as any occurrence of the following: myocardial infarction, acute

coronary syndrome not resulting in myocardial infarction, stroke, acute decompensated heart failure,

or death from cardiovascular causes. Serious adverse events are defined in SPRINT as any

occurrence of the following: emergency department evaluations for hypotension, syncope,

Page 38 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


bradycardia, electrolyte imbalance, injurious fall, or hospitalizations for acute kidney injury or acute

renal failure. Blood pressure is defined as the final measurement of SPRINT.

Exposure. Our exposure is patients’ average number of antihypertensive medications over the study

period.

Statistical analysis. To account for confounding by indication, we will perform an instrumental

variable analysis to assess the incremental (“marginal”) effect of adding an additional

antihypertensive medication on cardiovascular events and adverse events. Specifically, we will

instrument for patients’ mean exposure to antihypertensive medications over the study period by

whether they were randomized to the intensive or standard group. We will then stratify by the number

of antihypertensive medications at baseline to assess whether the marginal effect of antihypertensive

medications varies by whether it was the first, second, third, or fourth or more medication added to a

patient’s regimen.

We will estimate effects on cardiovascular and adverse events using Aalen additive hazard models to

account for the right-censored nature of clinical events. We will estimate effects on blood pressure

using two-stage least squares models. We will compare results from instrumental variable analyses to

results from multivariable regression models that do not account for endogeneity.

We will perform several sensitivity analyses to assess whether the three conditions for instrument

validity are met.

Condition #1: Random assignment. We will assess covariate balance across the intensive

(target < 120 mm Hg) versus standard (target < 140 mm Hg) groups.

Condition #2: Instrument strength. We will assess the correlation between the instrument

(randomization status) and the endogenous treatment (mean number of drugs over the study

period).

Condition #3: Exclusion restriction. We will assess whether randomization to the intensive vs.

standard group affects patient outcomes through mechanisms other than antihypertensive

regimen changes. These include nonpharmacologic interventions (more behavioral counseling

to intensive patients, resulting in greater changes in exercise, diet, alcohol consumption,

smoking, etc.) and pharmacologic interventions (increased adherence, increases dosages,

substitution for more powerful classes/agents, etc.). We will examine test for differences in the

following outcomes across intensive vs. standard groups: (A) differential nonpharmacologic

interventions: decreasing smoking status; altered lipid profile; decreased weight; and (B)

differential pharmacologic interventions: increased dosages; altered classes, improved

adherence (described by the Adherence Scale).

Page 39 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Results.

Table 1. Baseline characteristics of the SPRINT study participants

By number of antihypertensive drugs at baseline

Characteristics Total 0 1 2 3 4+

n

Age > 75

Female

Black

BMI > 35

Smoking status

Framingham 10-yr cardiovascular disease risk score

Previous cardiovascular disease

Previous CKD

Blood pressure

Figure 1. Incremental effect on major cardiovascular events

Forest plot consisting of the following:

• Models not adjusted for confounding o Overall effect o 0 drugs at baseline o 1 drug at baseline o 2 drugs at baseline o 3 drugs at baseline o 4+ drugs at baseline

• Instrumental variable models o Overall effect o 0 drugs at baseline o 1 drug at baseline o 2 drugs at baseline o 3 drugs at baseline o 4+ drugs at baseline

Figure 2. Incremental effect on serious adverse events: (same as Figure 1)

Figure 3. Incremental effect on systolic blood pressure: (same as Figure 1)

Page 40 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Table 2. Subgroup analysis of incremental effects on major cardiovascular events, serious

adverse events, and systolic blood pressure

Subgroup Major cardiovascular

events Serious adverse events Systolic blood pressure

Models not adjusted for confounding

Age

< 75 yr

≥ 75 yr

Sex

Female

Male

Race

Black

Nonblack

BMI

BMI < 35

BMI ≥ 35

Framingham risk score

FRS < ?

FRS ≥ ?

Previous CVD

Yes

No

Previous CKD

Yes

No

Instrumental variable analysis

(Repeat as above)

Page 41 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Major revisions to the original study plan Exposure.

Due to data limitations, we evaluated changes in the number of antihypertensive drug classes instead

of number of antihypertensive drug agents. Although the SPRINT team did collect data on number of

antihypertensive drug agents, drug class, and drug dosage, these data were not available to outside

researchers at the time of our analysis. As a result, we used number of drug classes both as our main

exposure and when stratifying by number of drug (classes) at baseline. When stratifying by number of

drug classes at baseline, we pooled patients who used four (rather than five) or more drug classes at

baseline to improve the precision of our estimates. Additionally, for our main exposure we selected

the number of drug classes at the study’s end, rather than the mean number of drug classes over the

study period.

Primary outcomes.

We included systolic blood pressure as a primary outcome in addition to major cardiovascular events

and major serious adverse events. Additionally, we shifted toward using blood pressure instead of

clinical events for our subgroup and sensitivity analyses. When creating our research plan, we were

uncertain about the number of clinical events that occurred in SPRINT. Upon receiving the data we

became aware that these events were sporadic enough that we were not powered to stratify

simultaneously by number of drugs at baseline and other covariates (e.g., gender, race). As a result,

we performed the majority of our secondary analyses using systolic blood pressure, an important

surrogate outcome.

Sensitivity analyses.

We were unable to perform several of our planned sensitivity analyses due to data limitations.

Specifically, we were unable to verify whether randomization status was associated with changes to

smoking status, weight, lipid profiles (i.e., nonpharmacologic interventions) or changes to

antihypertensive drug adherence, dosage, or specific classes (i.e., pharmacologic interventions other

than number of antihypertensive drug classes). In light of these data limitations, we assessed for

differential exposure to nonpharmacologic interventions by assessing whether incremental effects in

our main analysis were similar to those assessed: (1) at times we considered too early to be driven by

behavioral changes (i.e., 3-month visit); (2) among patients we considered unlikely to be targeted for

nonpharmacologic interventions (i.e., patients who were not obese at baseline, patients who did not

smoke at baseline).

Page 42 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Online Appendix Table 1. Inclusion and exclusion criteria (adapted from SPRINT analysis22)

Inclusion criteria 1. Age ≥ 50 years old AND 2. Systolic blood pressure (SBP):

a) 130-180 mm Hg on 0-1 antihypertensive drugs

b) 130-170 mm Hg on ≤ 2 antihypertensive drugs c) 130-160 mm Hg on ≤ 3 antihypertensive drugs d) 130-150 mm Hg on ≤ 4 antihypertensive drugs

“If a screenee is otherwise eligible for SPRINT but presents with a treated BP and/or number of medications that fall outside the SPRINT inclusion criteria, BP-lowering medications may be adjusted prior to the randomization visit to determine whether, with such adjustments, the screenee will meet eligibility criteria for SPRINT. A screenee who presents on no BP medications should have documentation of SBP ≥130 mm Hg on 2 visits within 3 months prior to the randomization visit in order to be eligible for the trial.”

AND 3. Cardiovascular risk (≥ 1 of the following)

a) Presence of clinical* or subclinical** cardiovascular disease other than stroke; OR b) CKD, defined as eGFR 20 – 59 ml/min/1.73m2 based on the 4-variable Modification of Diet in Renal Disease

(MDRD) equation and latest lab value, within the past 6 months. (If the serum creatinine is unstable within the last 6 months, enrollment into SPRINT could be delayed until the serum creatinine has been stabilized and the eGFR is still within the allowed range.); OR

c) Framingham Risk Score for 10-year CVD risk ≥ 15% based on laboratory work done within the past 12 months for lipids; OR

d) Age ≥ 75 years. * Clinical CVD (other than stroke)

a) Previous myocardial infarction (MI), percutaneous coronary intervention (PCI), coronary artery bypass grafting (CABG), carotid endarterectomy (CE), carotid stenting

b) Peripheral artery disease (PAD) with revascularization c) Acute coronary syndrome with or without resting ECG change, ECG changes on a graded exercise test (GXT), or

positive cardiac imaging study d) At least a 50% diameter stenosis of a coronary, carotid, or lower extremity artery e) Abdominal aortic aneurysm (AAA) ≥5 cm with or without repair

** Subclinical CVD a) Coronary artery calcium score ≥ 400 Agatston units within the past 2 years. b) Ankle brachial index (ABI) ≤0.90 within the past 2 years. c) Left ventricular hypertrophy (LVH) by ECG (based on computer reading), echocardiogram

Exclusion criteria 1. An indication for a specific BP lowering medication (e.g., beta-blocker following acute myocardial infarction) that the

person is not taking and the person has not been documented to be intolerant of the medication class. (If a screenee has a non-hypertension indication for a BP-lowering medication (e.g., beta blocker post-MI, renin angiotensin system (RAS) blocker for CVD prevention, or alpha blocker for benign prostatic hypertrophy (BPH)), the screenee should be on the appropriate dose of such medication before assessing whether he/she meets the SPRINT inclusion criteria. If the investigator believes that a potential participant has such an indication but is not receiving appropriate treatment, he/she should encourage the potential participant’s primary care provider to consider placing the patient on the appropriate therapy prior to proceeding with the screening process.)

OR

Page 43 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


2. Known secondary cause of hypertension that causes concern regarding safety of the protocol. OR 3. One minute standing SBP < 110 mm Hg. Not applicable if unable to stand due to wheelchair use. OR 4. Proteinuria in the following ranges (based on a measurement within the past 6 months)

a. 24 hour urinary protein excretion ≥1 g/day, or b. If measurement (a) is not available, then 24 hour urinary albumin excretion ≥ 600 mg/day, or c. If measurements (a) or (b) are not available, then spot urine protein/creatinine ratio ≥ 1 g/g creatinine, or d. If measurements (a), (b), or (c) are not available, then spot urine albumin/creatinine ratio ≥ 600 mg/g

creatinine, or e. If measurements (a), (b), (c), or (d) are not available, then urine dipstick 2+ protein

OR 5. Arm circumference too large or small to allow accurate blood pressure measurement with available devices OR 6. Diabetes mellitus. Participants taking medications for diabetes at any time in the last 12 months are excluded.

Participants are also excluded if there is documentation of: FPG at or above 126 mg/dL, A1C ≥6.5 percent, a two-hour value in an OGTT (2-h PG) at or above 200 mg/dL or a random plasma glucose concentration ≥200 mg/dL. The diagnosis of diabetes must be confirmed on a subsequent day by repeat measurement, repeating the same test for confirmation. However, if two different tests (eg, FPG and A1C) are available and are concordant for the diagnosis of diabetes, additional testing is not needed. If two different tests are discordant, the test that is diagnostic of diabetes should be repeated to confirm the diagnosis.

OR 7. History of stroke (not CE or stenting) OR 8. Diagnosis of polycystic kidney disease OR 9. Glomerulonephritis treated with or likely to be treated with immunosuppressive therapy OR 10. eGFR < 20 ml/min /1.73m2 or end-stage renal disease (ESRD) OR 11. Cardiovascular event or procedure (as defined above as clinical CVD for study entry) or hospitalization for unstable

angina within last 3 months OR 12. Symptomatic heart failure within the past 6 months or left ventricular ejection fraction (by any method) < 35% OR

Page 44 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


13. A medical condition likely to limit survival to less than 3 years, or a cancer diagnosed and treated within the past two

years that, in the judgment of clinical study staff, would compromise a participant’s ability to comply with the protocol and complete the trial. Exceptions to the exclusion for diagnosed cancer would include, for example, non-melanoma skin cancer, early-stage prostate cancer, localized breast cancer.

OR 14. Any factors judged by the clinic team to be likely to limit adherence to interventions. For example

a. Active alcohol or substance abuse within the last 12 months b. Plans to move outside the clinic catchment area in the next 2 years without the ability to transfer to another

SPRINT site, or plans to be out of the study area for more than 3 months in the year following enrollment. c. Significant history of poor compliance with medications or attendance at clinic visits d. Significant concerns about participation in the study from spouse, significant other, or family members e. Lack of support from primary health care provider f. Residence too far from the study clinic site such that transportation is a barrier including persons who require

transportation assistance provided by the SPRINT clinic funds for screening or randomization visits g. Residence in a nursing home. Persons residing in an assisted living or retirement community are eligible if

they meet the other criteria. h. Clinical diagnosis of dementia, treatment with medications for dementia, or in the judgment of the clinician

cognitively unable to follow the protocol i. Other medical, psychiatric, or behavioral factors that in the judgment of the Principal Investigator may

interfere with study participation or the ability to follow the intervention protocol OR 15. Failure to obtain informed consent from participant OR 16. Currently participating in another clinical trial (intervention study). Note: Patient must wait until the completion of

his/her activities or the completion of the other trial before being screened for SPRINT. OR 17. Living in the same household as an already randomized SPRINT participant OR 18. Any organ transplant OR 19. Unintentional weight loss > 10% in last 6 months OR 20. Pregnancy, currently trying to become pregnant, or of child-bearing potential and not using birth control

Page 45 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Online Appendix Methods

Estimating differences in incremental effects on systolic blood pressure across

baseline drug use and patient subgroups. For standard multivariable adjusted

models, we implemented interaction models to formally test whether incremental effects

of antihypertensive drugs varied systematically across baseline number of drug classes.

Specifically, we included an interaction between the baseline number and final number

of antihypertensive drug classes. We did not estimate similar interaction models for

instrumental variable models because interacting the instrument (randomization status)

with a potential confounder (the number of drug classes at baseline) would render the

instrumental variable analysis invalid.

For instrumental variable models, we instead tested for trends in the incremental effects

by testing for differences in the incremental effects of adding a: (a) first versus second

drug class; (b) second versus third drug class; (c) fourth versus third class; and (d)

fourth or more versus a first drug class. To do this, we estimated separate models

stratified by baseline number of drug classes and tested for differences in effect

estimates across baseline strata in the following four steps: (1) estimated the two-stage

least-squares regression for patients in the first stratum (e.g., using zero drug classes at

baseline) “by hand”, i.e., estimating the effect of the predicted number of

antihypertensive drug classes from the first stage as a function of randomization status

and covariates; (2) estimated the two-stage least-squares regression for patients in the

second stratum (e.g., using one drug class at baseline), again by hand; (3) combined

Page 46 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


the parameter estimates and associated (co)variance from the two models (using

Stata’s “suest” command); (4) evaluated whether the incremental effects differed

between the strata by testing whether linear combinations of the two parameter

estimates differed from zero).

In these analyses (reported in online appendix table 9), differences in the incremental

effects across baseline strata can be interpreted in the following manner: ∆ < 0 mm Hg

suggests a synergistic benefit (i.e., relatively greater reductions in blood pressure); ∆ >

0 mm Hg suggests a diminishing benefit; and ∆ = 0 mm Hg suggests an additive

benefit. In online appendix table 9, all models failed to reject the null hypothesis that

incremental effects varied across baseline strata, suggesting that incremental effects on

systolic blood pressure are approximately additive (vs. diminishing or synergistic), i.e.,

similar with each added drug. This is in accordance with Figure 1, which demonstrates a

relatively stable effect of estimate of approximately 14-15 mm Hg drop in SBP with each

added drug. For example, where Figure 1 displays the incremental effect of adding the

1st drug (-13.9 mm Hg) and the effect of adding the 2nd drug (-14.2 mm Hg), online

Table 9 shows that the difference between adding the 2nd drug and 1st drug is -14.2 -

(-13.9) = -0.3 (95% -2.2, 1.5) mm Hg. Furthermore, because estimating instrumental

variable estimates by hand does not account for the combined statistical uncertainty of

both the first- and second-stages, this procedure results in overly precise parameter

estimates and is thus more likely to falsely reject the null hypothesis that incremental

effects do not differ across baseline strata (Type I error). Thus, our failure to reject the

Page 47 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


null hypothesis further confirms our conclusion that the effects of antihypertensive drugs

are additive (i.e., do not diminish) with each added drug.

We followed a similar procedure for standard multivariable adjusted models, now

estimating blood pressure changes as a function of observed rather than predicted

antihypertensive drug use. We also repeated this procedure to test for differences in the

incremental effects of antihypertensive drugs across patient subgroups (e.g., difference

in the incremental effect of adding the 1st drug between men and women).

Estimating incremental effects on major cardiovascular events and serious

adverse events. We estimated a recently validated two-stage additive hazards model

to examine the incremental effect of antihypertensive drugs on major cardiovascular

events and serious adverse events.2 In the first stage, we estimated the predicted

number of antihypertensive drug classes as a linear function of randomization status

and covariates. In the second stage, we estimated cardiovascular or adverse risk as a

function of the predicted number of antihypertensive drug classes (from the first stage)

and covariates. We estimated Aalen additive hazards models in the second stage to

account for the right-censored nature of survival outcomes. To account for the

combined statistical uncertainty of the two stages, we implemented a bootstrap

estimator and based statistical inference on 95% confidence intervals derived from 2000

nonparametric bootstraps.

Page 48 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


We specifically used additive hazards models rather than proportional hazards models

(e.g., Cox proportional hazards) because prior work has demonstrated that estimating a

two-stage proportional hazards model is valid only when the outcome is “rare” (i.e., with

survival approaching unity),2 an assumption that is not met in the present study. We

estimated fully non-parametric additive hazards models and based parameter estimates

on the weighted linear regression of the cumulative estimates plot. We did so to provide

a useful measure of the overall size of the effect.

We performed three sensitivity tests to evaluate the assumptions of the additive hazards

models, namely that the effects of antihypertensive drugs and other factors are additive

and constant over time (i.e., time-invariant.). First, we plotted observed cumulative

estimates over time of the effects of antihypertensive drugs and other covariates on

composite major cardiovascular events. Second, we formally tested whether effects

were additive and constant using the Kolmogorov-Smirnov test, which is a

nonparametric goodness-of-fit test that assesses the degree to which the observed

cumulative distribution function (CDF) fits a hypothetical CDF based on the assumption

of time-invariant additive effects. Finally, we compared estimates from semi-parametric

time-invariant additive hazards models to pooled estimates from non-parametric time-

varying additive hazards models. The results of these sensitivity analyses are given in

online appendix table 8 and figure 7.

Page 49 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Online Appendix Figure 1. CONSORT diagram

Page 50 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Online Appendix Table 2. Distribution of number of antihypertensive drug classes at each

patient’s final visit

Final number of antihypertensive drug classes

a

Standard (N= 4523) Intensive (N= 4569) P value

0 486 (10.7%) 80 (1.8%) <0.001

1 1432 (31.7%) 464 (10.2%)

2 1505 (33.3%) 1383 (30.3%)

3 796 (17.6%) 1499 (32.8%)

4 250 (5.5%) 855 (18.7%)

5 50 (1.1%) 242 (5.3%)

6 3 (0.1%) 39 (0.9%)

7 1 (<1%) 7 (0.2%)

a Final number of antihypertensive drug classes were measured at the latest visit for which there was

also a recorded blood pressure measurement.

Page 51 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Online Appendix Figure 2. Change in number of antihypertensive drug classes over study

Page 52 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Online Appendix Table 3. Change in number of antihypertensive drug classes over study

period by number of drug classes at baseline

Change in number of drug classes over study perioda

Number of antihypertensive drug classes

at baseline -4 -3 -2 -1 0 +1 +2 +3 +4 +5

Standard group (target < 140 mm Hg)

Overall 0 0 5 74 224 111 19 3 0 1

1 0 0 26 287 673 301 43 4 0 0

2 0 5 96 424 718 275 47 7 0 0

3 or more 3 27 133 392 414 179 24 2 0 0

Intensive group (target < 120 mm Hg)

Overall 0 0 4 26 129 169 74 16 3 0

1 0 0 19 100 466 507 180 48 5 0

2 0 5 30 168 652 537 191 36 5 1

3 or more 4 9 39 170 504 358 93 13 3 1 aRepresents change between number of drug classes at the baseline visit and the final visit of the study at which there

was also a blood pressure measurement. Numbers represent counts of patients in each cell

Page 53 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Online Appendix Table 4. Instrument strength stratified by number of drug classes at baseline

Number of drug classes at baseline

F statistic Partial R2

0 230 0.26 1 515 0.23 2 463 0.19 3 or more 298 0.15

Page 54 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Online Appendix Table 5. Baseline characteristics of the SPRINT study participants by SPRINT randomization status

Characteristics Standard (N= 4523) Intensive (N= 4569) P value

Age - years (SD) 67.8 (9.4) 67.9 (9.4) 0.75

Female - no. (%) 1583 (35.0%) 1634 (35.8%) 0.45

Black - no. (%)a

1442 (31.9%) 1414 (30.9%) 0.34

BMI - kg/m2 (SD)

b 29.8 (5.7) 29.9 (5.8) 0.36

Fasting HDL - mg/dl (SD) 52.7 (14.6) 52.9 (14.4) 0.58

Serum creatinine - mg/dl (SD) 1.1 (0.3) 1.1 (0.3) 0.87

Statin use - no. (%) 2018 (44.6%) 1952 (42.7%) 0.069

Ever smoker - no. (%) 2530 (55.9%) 2566 (56.2%) 0.83

History of cardiovascular disease - no. (%) 760 (16.8%) 764 (16.7%) 0.92

History of chronic kidney disease - no. (%)c

1267 (28.0%) 1305 (28.6%) 0.56

Blood pressure - mm Hg (SD)

Systolic 139.7 (15.4) 139.7 (15.8) 0.95

Diastolic 78.1 (12.0) 78.2 (11.9) 0.57

SI conversions factor: To convert the values for cholesterol to millimoles per liter, multiply by 0.02586. To convert the values for glucose to millimoles per liter, multiply by 0.05551. HDL denotes high-density lipoprotein. aRace was self-report. Black race includes Hispanic black and black.

bThe body-mass index is the weight in kilograms divided by the square of the height in meters.

cNo history of chronic kidney disease includes some participants with unknown chronic kidney disease status at

baseline.

Page 55 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Online Appendix Figure 3. Incremental effects across clinical and demographic subgroups Online Appendix Figure 3A. Age

Diamonds represent point estimates from pooled models. Squares represent point estimates from models stratified by baseline number of drug classes. Lines represent 95% confidence intervals. Antihypertensive drug classes are measured at baseline and at each patient’s final visit. Instrumental variable models were estimated using two-stage ordinary least squares regression. Online Appendix Figure 3C. Race was self-reported. Black race includes non-Hispanic and Hispanic black participants. Online Appendix Figure 3E. Obesity is defined as a body-mass index greater than or equal to 35. Online Appendix Figure 3F. CVD is cardiovascular disease. Online Appendix Figure 3G. CKD is chronic kidney disease. No history of CKD includes some participants with unknown CKD status at baseline.

Page 56 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Online Appendix Figure 3B. Sex

Page 57 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Online Appendix Figure 3C. Black race

Page 58 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Online Appendix Figure 3D. Smoking status

Page 59 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Online Appendix Figure 3E. Obesity

Page 60 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Online Appendix Figure 3F. History of cardiovascular disease

Page 61 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Online Appendix Figure 3G. History of chronic kidney disease

Page 62 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Online Appendix Table 6. Incremental effects of antihypertensive drugs on systolic blood pressure, major cardiovascular events, and serious adverse events at three-month visit and over the study period

Systolic blood pressure, (mm Hg)

Major cardiovascular events (per thousand person-years)

Serious adverse events (per thousand person-years)

Incremental effects on blood pressure (95% CI)

Models not adjusted for confounding

Number of drug classes at three-month visit

a -1.0 (-1.4, -0.7) 2.7 (0.4, 5.2) 17.4 (9, 26.1)

Mean number of drug classes over study period

b -2.0 (-2.3, -1.6) 2.3 (0, 4.5) 13.7 (5.7, 22)


Number of drug classes at three-month visit

a -20.9 (-22.5, -19.2) -9.1 (-16.2, -2.3) 18.3 (-6, 43.6)

Mean number of drug classes over study periodb

-16.4 (-17.5, -15.3) -7.1 (-12.8, -1.7) 14.2 (-5.4, 33.4)

aSystolic blood pressure is measured at the three-month visit. Antihypertensive drug classes are measured at

baseline and at the exit of the two-month visit. bSystolic blood pressure is measured at the final visit. We calculated the mean number of antihypertensive drug

classes a patient was recorded as being prescribed over the study period. CI is confidence interval.

Page 63 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Online Appendix Figure 4. Incremental effect of antihypertensive drugs on systolic blood pressure at the three-month visit

Diamonds represent point estimates from pooled models. Squares represent point estimates from models stratified by baseline number of drug classes. Lines represent 95% confidence intervals. Systolic blood pressure is measured at the three-month visit. Antihypertensive drug classes are measured at baseline and at exit of the two-month visit. Multivariable adjusted models were estimated using ordinary least squares regression. Instrumental variable models were estimated using two-stage ordinary least squares regression.

Page 64 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Online Appendix Figure 5. Incremental effect of mean number of antihypertensive drugs on systolic blood pressure

Diamonds represent point estimates from pooled models. Squares represent point estimates from models stratified by baseline number of drug classes. Lines represent 95% confidence intervals. Systolic blood pressure is measured at the final visit. We calculated the mean number of antihypertensive drug classes a patient was recorded as being prescribed over the study period. Multivariable adjusted models were estimated using ordinary least squares regression. Instrumental variable models were estimated using two-stage ordinary least squares regression.

Page 65 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Online Appendix Table 7. Incremental effects of antihypertensive drugs on total number of blood pressure measurements

Incremental effects on number of blood pressure measurements (95% confidence interval)

Multivariable adjusted models

Final number of drug classes 0.15 (0.07, 0.23)

Mean number of drug classes over study period 0.17 (0.08, 0.25)


Final number of drug classes 0.07 (-0.09, 0.22)

Mean number of drug classes over study period 0.07 (-0.10, 0.25)

Page 66 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Online Appendix Table 8. Nonparametric tests of constant additive effects on composite major cardiovascular events and serious adverse events

P valuea

Estimate Major cardiovascular events

Serious adverse events

Antihypertensive drug 0.334 0.435 Age 0.735 0.079 History of cardiovascular disease 0.876 0.568 History of chronic kidney diseaseb 0.132 0.349 Creatinine 0.098 0.430 Female 0.564 0.354 Blackc 0.896 0.401 BMI 0.336 0.383 HDL 0.762 0.324 Smoker 0.588 0.080 Statin 0.242 0.665 Intercept 0.727 0.136 aKolmogorov-Smirnov test: null hypothesis is constant (time-invariant) additive

effect. bNo history of chronic kidney disease includes some participants with unknown

chronic kidney disease status at baseline. cRace was self-report. Black race includes non-Hispanic and Hispanic black

participants.

Comparison with semi-parametric additive hazards models. We estimated semi-parametric additive hazards models and confirmed that time-invariant parameters did not differ significantly from coefficients derived from weighted averages of time-variant, non-parametric models. In semi-parametric models, the incremental effect of antihypertensive drug classes was -7.7 events per 1,000 patient years for composite major cardiovascular events (95% CI, -16.2 to 0.8) and 10.2 events per 1,000 patient years for composite serious adverse events (95% CI, 0.8 to 20.0).

Page 67 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Online Appendix Figure 6. Cumulative coefficient plots of the incremental effects

on composite major cardiovascular events

Page 68 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Page 69 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Page 70 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Page 71 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Page 72 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Page 73 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Online Appendix Table 9. Differences in incremental effects on blood pressure when adding first, second, third, or fourth or more antihypertensive drug class

Difference in incremental effects on blood pressure across baseline number of drug classes

2nd vs. 1st 3rd vs. 2nd 4th vs. 3rd 4th vs. 1st

Difference

(95% CI) P value

Difference

(95% CI) P value

Difference

(95% CI) P value

Difference

(95% CI) P value

Interaction

terma

P value for

interactiona

Standard adjusted models

Systolic 0.5 (-0.5, 1.4) 0.348 0.3 (-0.4, 1.0) 0.461 0.9 (0.1, 1.6) 0.033 1.6 (0.6, 2.6) 0.002 0.6 <0.001

Diastolic 0.3 (-0.5, 1.0) 0.488 0.0 (-0.4, 0.5) 0.864 0.4 (-0.1, 0.9) 0.135 0.7 (-0.1, 1.4) 0.068 0.2 0.024

Instrumental variable modelsd

Systolic -0.3 (-2.2, 1.5) 0.734 -0.5 (-1.9, 0.9) 0.463 -0.4 (-2.0, 1.3) 0.666 -1.2 (-3.3, 0.8) 0.247 --b --

b

Diastolic 0.3 (-1.2, 1.7) 0.723 -0.7 (-1.7, 0.4) 0.200 0.2 (-1.0, 1.4) 0.766 -0.2 (-1.8, 1.3) 0.775 --b --

b

aWe interacted the final number of antihypertensive drug classes with the number of drug classes at baseline to formally test whether the incremental effects of

antihypertensive drugs varied systematically across baseline number of drug classes. bWe did not estimate interaction models for our instrumental variable analysis because interacting the instrument (randomization status) with a potential confounder

(the number of drug classes at baseline) would render the instrumental variable analysis invalid. CI is confidence interval.

Page 74 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Online Appendix Figure 7. Incremental effect of antihypertensive drugs on diastolic blood pressure



Diastolic blood pressure represents each patient’s final recorded blood pressure measurement.

Antihypertensive drug classes are measured at baseline and at the latest visit at which there is also a

recorded blood pressure measurement. Multivariable adjusted models were estimated using ordinary

least squares regression. Instrumental variable models were estimated using two-stage ordinary least

squares regression.

Page 75 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Online Appendix Figure 8. Incremental effect of antihypertensive drugs on component major cardiovascular events and all-cause mortality

Diamonds represent point estimates from pooled models. Squares represent point estimates from models stratified by baseline number of drug classes. Lines represent 95% confidence intervals. Antihypertensive drug classes are measured at baseline and at each patient’s final visit. For patients who experienced a major cardiovascular event, we used the last recorded value of drug classes before event incidence. We estimated additive hazards models to account for the right-censored nature of survival outcomes such as risk of major cardiovascular events.28

Page 76 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Online Appendix Figure 9. Incremental effect of antihypertensive drugs on component serious adverse events

Diamonds represent point estimates from pooled models. Squares represent point estimates from models stratified by baseline number of drug classes. Lines represent 95% confidence intervals. Antihypertensive drug classes are measured at baseline and at each patient’s final visit. For patients who experienced a serious adverse event, we used the last recorded value of drug classes before event incidence. We estimated additive hazards models to account for the right-censored nature of survival outcomes such as risk of serious adverse events.28

Page 77 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Online Appendix Table 10. Incremental effects on systolic blood pressure across clinical and demographic subgroups

Adding 1st

drug class Adding 2nd

drug class Adding 3rd

drug class Adding 4

th or more drug

class

Subgroup Incremental effect (95% CI)

P value

Incremental effect (95% CI)

P value


P value


P value

Age < 75 years -14.7 (-16.0, -13.4) <0.001 -14.8 (-17.3, -12.4) <0.001 -14.6 (-16.6, -12.7) <0.001 -15.1 (-16.9, -13.2) <0.001 ≥ 75 years -13.7 (-15.7, -11.7) <0.001 -9.3 (-15.8, -2.8) 0.005 -13.1 (-16.1, -10.2) <0.001 -14.1 (-17.3, -10.9) <0.001 Difference

0.5 (-1.2, 2.3) 0.537 5.5 (0.7, 10.3) 0.023 1.5 (-1.0, 4.1) 0.248 1.0 (-1.5, 3.5) 0.440

Sex Male -13.7 (-14.9, -12.6) <0.001 -12.7 (-15.2, -10.3) <0.001 -13.0 (-14.4, -11.5) <0.001 -14.6 (-16.5, -12.8) <0.001 Female -15.8 (-17.8, -13.8) <0.001 -18.8 (-26.8, -10.8) <0.001 -16.7 (-20.5, -13.0) <0.001 -14.8 (-17.6, -12.0) <0.001 Difference 0.5 (-1.2, 2.3) 0.537 -6.1 (-10.9, -1.3) 0.013 -3.8 (-6.0, -1.5) 0.001 -0.2 (-2.4, 2.1) 0.886

Racea

Non-black -14.6 (-15.9, -13.3) <0.001 -14.6 (-17.6, -11.6) <0.001 -13.6 (-15.2, -11.9) <0.001 -14.7 (-16.8, -12.6) <0.001 Black -14.1 (-15.8, -12.4) <0.001 -11.5 (-15.3, -7.6) <0.001 -16.1 (-20.1, -12.1) <0.001 -14.6 (-17.3, -11.9) <0.001 Difference 0.5 (-1.2, 2.3) 0.537 3.1 (-0.2, 6.5) 0.064 -2.5 (-4.8, -0.2) 0.036 0.1 (-2.1, 2.3) 0.927

Obesitya

Non-obese -14.5 (-15.6, -13.3) <0.001 -13.5 (-16.0, -11.1) <0.001 -14.4 (-16.2, -12.6) <0.001 -15.3 (-17.2, -13.4) <0.001 Obese -14.0 (-16.4, -11.5) <0.001 -16.0 (-24.1, -7.8) <0.001 -11.8 (-16.4, -7.3) <0.001 -12.2 (-15.2, -9.2) <0.001 Difference 0.5 (-1.2, 2.3) 0.537 -2.4 (-6.7, 1.9) 0.271 2.6 (-0.5, 5.6) 0.096 3.1 (0.5, 5.7) 0.019

Smoking Never-smoker

-14.6 (-16.3, -12.9) <0.001 -11.8 (-15.2, -8.3) <0.001 -14.2 (-16.7, -11.8) <0.001 -15.9 (-18.7, -13.0) <0.001

Ever-smoker -14.3 (-15.5, -13.0) <0.001 -15.4 (-18.5, -12.4) <0.001 -14.2 (-16.2, -12.1) <0.001 -14.0 (-15.8, -12.1) <0.001 Difference 0.5 (-1.2, 2.3) 0.537 -3.7 (-6.8, -0.5) 0.022 0.0 (-1.9, 2.0) 0.961 1.9 (-0.2, 4.0) 0.075

History of cardiovascular disease No -14.5 (-15.8, -13.2) <0.001 -13.9 (-16.3, -11.5) <0.001 -14.6 (-16.5, -12.6) <0.001 -14.9 (-16.7, -13.1) <0.001 Yes -14.0 (-16.4, -11.5) <0.001 -13.0 (-19.0, -6.9) <0.001 -12.3 (-16.3, -8.3) <0.001 -13.7 (-17.6, -9.9) <0.001 Difference 0.5 (-1.2, 2.3) 0.537 0.9 (-3.5, 5.4) 0.685 2.2 (-0.7, 5.1) 0.135 1.2 (-1.4, 3.7) 0.369

History of chronic kidney diseasec

No -14.4 (-15.6, -13.2) <0.001 -13.8 (-15.8, -11.8) <0.001 -14.8 (-16.8, -12.8) <0.001 -15.0 (-16.8, -13.2) <0.001 Yes -14.6 (-16.9, -12.3) <0.001 -18.3 (-33.3, -3.3) 0.017 -12.5 (-15.6, -9.4) <0.001 -14.3 (-17.5, -11.1) <0.001 Difference 0.5 (-1.2, 2.3) 0.537 -4.5 (-13.1, 4.1) 0.301 2.3 (-0.1, 4.8) 0.061 0.7 (-1.6, 3.1) 0.551

aRace was self-report. Black race includes non-Hispanic and Hispanic black participants.

bObesity is defined as a body-mass index greater than or equal to 35.

cNo history of chronic kidney disease includes some participants with unknown chronic kidney disease status at baseline.

CI is confidence interval.

Page 78 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Online Appendix Figure 10. Incremental effect of antihypertensive drugs on systolic blood pressure in unadjusted models

Diamonds represent point estimates from pooled models. Squares represent point estimates from models stratified by baseline number of drug classes. Lines represent 95% confidence intervals. Systolic blood pressure represents each patient’s final recorded blood pressure measurement. Antihypertensive drug classes are measured at baseline and at the latest visit for which there was also a recorded blood pressure measurement. Unadjusted bivariate models were estimated using ordinary least squares and were not adjusted for any covariates. Unadjusted instrumental variable models were estimated using two-stage ordinary least squares regression and were not adjusted for any covariates.

Page 79 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Online Appendix Figure 11. Incremental effect of adding a fifth or more antihypertensive drug class on systolic blood pressure

Diamonds represent point estimates from pooled models. Squares represent point estimates from models stratified by baseline number of drug classes. Lines represent 95% confidence intervals. Systolic blood pressure represents each patient’s final recorded blood pressure measurement. Antihypertensive drug classes are measured at baseline and at the latest visit for which there was also a recorded blood pressure measurement. Multivariable adjusted models were estimated using ordinary least squares regression. Instrumental variable models were estimated using two-stage ordinary least squares regression.

Page 80 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Online Appendix Table 11. Intention-to-treat analyses of the effect of randomization status on changes in prescribed number of drug classes and systolic blood pressure

Change in number of drug classes

Change in systolic blood pressure (mm Hg)

Overall effect of randomization status 1.0 (1.0, 1.0) -14.1 (-13.4, -13.4)

By number of antihypertensive drug classes at baseline

Effect among patients on zero drug classes at baseline 1.1 (1.2, 1.2) -15.3 (-13.6, -13.6)

Effect among patients on 1 drug class at baseline 1.0 (1.1, 1.1) -14.3 (-13.2, -13.2)

Effect among patients on 2 drug classes at baseline 0.9 (1.0, 1.0) -14.0 (-12.9, -12.9)

Effect among patients on 3 or more drug classes at baseline 0.9 (1.0, 1.0) -13.6 (-12.3, -12.3)

Page 81 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Online Appendix References 1. Group SR, Wright JT, Williamson JD, et al. A Randomized Trial of Intensive versus Standard

Blood-Pressure Control. N Engl J Med. 2015;373(22):2103-2116. http://eutils.ncbi.nlm.nih.gov/entrez/eutils/elink.fcgi?dbfrom=pubmed&id=26551272&retmode=ref&cmd=prlinks.

2. Tchetgen Tchetgen EJ, Walter S, Vansteelandt S, Martinussen T, Glymour M. Instrumental Variable Estimation in a Survival Context. Epidemiology. 2015;26(3):402-410. http://content.wkhealth.com/linkback/openurl?sid=WKPTLP:landingpage&an=00001648-201505000-00016.

3. Aalen OO. A linear regression model for the analysis of life times. Stat Med. 1989;8(8):907-925. http://doi.wiley.com/10.1002/sim.4780080803.

Page 82 of 82


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

confidential: for review only · confidential: for review only the incremental effects of...

Documents