Initiation of Antihypertensive Medication from Midlife on Incident Dementia: The Health and Retirement Study

INTRODUCTION

Midlife hypertension has been identified as one of the most important modifiable risk factors of dementia [1]. Evidence from prospective cohort studies has indicated that high blood pressure in midlife is associated with a higher risk of cognitive decline and dementia [2–5]. It is therefore hypothesized that reducing blood pressure in midlife may prevent incident dementia in populations entering late life. Although findings from observational studies showed that use of antihypertensive medication, the major approach to reducing blood pressure, is associated with a lower risk of dementia [6], such treatment effect has been rarely demostrated in randomized controlled trials (RCTs), which would provide a higher level of evidence by design and principle. In a recently published systematic review and meta-analysis [7], although it concluded that antihypertensive medication use is significantly associated with a lower risk of incident dementia, only one out of twelve RCTs showed significant results [8], while none of the other RCTs showed a significant reduction in risk of dementia from antihypertensive medication use [9–14].

Efforts have been made to explain the gap in contradictory findings between RCTs and observational studies. One of the potential explanations is that while high blood pressure in midlife is associated with a higher risk of dementia, high blood pressure in late life does not predict the risk of dementia well, as is documented in the existing literature [15–20]. Cohort studies have found that antihypertensive medication is associated with a lower risk of incident dementia among older adults. However, there is potential for confounding due to the possibility that antihypertensive medication was initiated in earlier stages of life. In addition, in observational studies, other factors that affect medication use and dementia risk may not be fully adjusted in the analysis. Therefore, initiating systolic blood pressure in late life may not lead to a lower risk of dementia. However, except for the SPRINT-MIND trial (Systolic Blood Pressure Intervention Trial - Memory and Cognition in Decreased Hypertension), which includes participants 50 years or older, all existing RCTs of antihypertensive medication on incident dementia are based on older adults. Thus, since midlife high blood pressure is more predictive of incident dementia, the target for intervention should be on blood pressure in midlife.

Ideally, an RCT should be conducted to estimate the long-term effectiveness of antihypertensive medication use in midlife on the prevention of dementia in late life. However, it would not be feasible, if not impossible in practice to enroll participants in midlife into a trial that lasts for decades. Moreover, given the huge burden of public health brought by dementia, it is not appropriate to spend decades for the results of a potential trial with numerous decision to be made during this long period of time. Given the fact that no existing trial started an intervention of antihypertensive medication use from midlife, evidence from such RCTs is unlikely to emerge in near future. Under such circumstances, the optimal way to draw a causal conclusion on the effect of antihypertensive medication use for preventing incident dementia in late life is to emulate a target trial with existing observational studies, in which causal questions are specified, and observational data is analyzed like RCTs.

Therefore, we aimed to use a prospective cohort study to emulate a target trial and estimate the effectiveness of initiation and sustained use of antihypertensive medication from midlife among hypertensive patients on preventing incident dementia in late life.

MATERIALS AND METHODS

RESULTS

A total of 2,375 subjects were included in the emulated target trial after applying all inclusion and exclusion criteria (Supplementary Figure 1), with 1,882 initiators and 493 noninitiators. The baseline characteristics of the participants in the emulated target trial were shown in Table 2. The initiators had a higher proportion of female participants, higher prevalence of diabetes, heart disease, and stroke. With application of IPTW and IPCW, the variables were balanced between two groups in the emulated target trial.

In the emulated target trial, during 22-year (mean: 14.1±7.5) follow-up period, a total of 309 (13.0%) dementia cases occurred. With adjustment for all covariates and weighted by IPTW and IPCW, by the end of 22-year follow-up, the risk of incident dementia was 23.0% and 29.4% among initiators and non-initiators, respectively. The initiation of antihypertensive medication showed a significant effect in reducing incident dementia after 18 years of follow-up (RR = 0.70, 95% CI: 0.51, 1.00). At the end of the follow-up period, initiating antihypertensive medication was found to reduce the risk of incident dementia by 22% (RR = 0.78, 95% CI: 0.63, 0.99) (Fig. 1, Table 3).

OPEN IN VIEWER

Fig. 1

Estimated cumulative incidence of dementia among participants who initiated antihypertensive medication and not initiated antihypertensive medication in the emulated target trial using pooled logistic regression models. The models were weighted for inverse-probability treatment weight and inverse-probability censoring weight based on baseline covariates, including age, sex, race, education, smoking status, physical activity, depressive symptoms, body mass index, prevalent diabetes, heart disease, and stroke.

After censoring participants as they deviated from the initial assignment of treatment at baseline, the estimated cumulative incidence of dementia estimated by pooled logistic regression models with IPTW and IPCW by the end of 22-year follow-up was 22.3% and 28.3% among sustained use of antihypertensive medication and persistent non-initiators, respectively. However, no significant effect was shown with adherence to antihypertensive medication on reducing incident dementia (Fig. 2, Table 4).

OPEN IN VIEWER

Fig. 2

Estimated cumulative incidence of dementia among participants who had sustained use of antihypertensive medication and those who never initiated antihypertensive medication in the emulated target trial using pooled logistic regression models. The models were weighted for inverse-probability treatment weight (base on baseline covariates) and inverse-probability censoring weight (based on baseline covariates (including age, sex, race, education, body mass index, smoking, depressive symptoms, physical activity, prevalent diabetes, heart disease, stroke) and time-varying covariates (smoking status, physical activity, depressive symptoms, body mass index, prevalent diabetes, heart disease, and stroke)).

For sensitivity, initiating antihypertensive medication use reduced the risk of dementia and death after 22 years of follow-up (RR = 0.90, 95% CI: 0.83, 0.97) (Supplementary Figure 2, Supplementary Table 1), and no significant reduction was found for sustained antihypertensive medication use on dementia and death (Supplementary Figure 3, Supplementary Table 2).

DISCUSSION

In an emulated target trial of antihypertensive medication use on reducing incident dementia starting from midlife, using the new-user design and application of IPTW and IPCW, we found that individuals with hypertension those who initiated antihypertensive medication use had a lower risk of dementia compared with those who did not, while we did not find effectiveness for sustained antihypertensive medication use on reduction of incident dementia.

In addition to multiple RCTs, previous efforts have been taken to draw a causal relationship between initiation of antihypertensive medication use and incident dementia using emulated target trials. Using the Rotterdam Study, Rojas-Saunero et al. studied the effects of hypothetical interventions on systolic blood pressure to reduce dementia with 15 years of follow-up [31]. The study defined multiple strategies for blood pressure reduction, including maintaining systolic blood pressure below 120 mmHg, and maintaining systolic blood pressure below 140 mmHg. However, compared to the natural course, blood pressure reduction did not achieve a reduction of incident dementia. There are several reasons that may explain the null results. First of all, consistency is an important issue, since the cause for different levels of blood pressure reduction is unknown, and a lower blood pressure may be achieved due to pathological (but unmeasured) reasons, such as hypotension. Improvement of lifestyle factors (e.g., physical activity, healthy diet) also reduces blood pressure levels. Therefore, there could be different multiple versions of treatment in the same group of people with the same level of blood pressure reduction. In the presence of multiple versions of treatments in the same group, there could be unmeasured confounders between the treated blood pressure during follow-up and incident dementia, and when the follow-up dementia is controlled, then the unmeasured confounder can bias the relationship between blood pressure reduction and incident dementia, which is known as collider bias [32, 33]. Also, all the participants included in the analysis were 65 years or older, and as is indicated in the literature, high blood pressure in late life may not be predictive of incident dementia, and late treatment of blood pressure may have missed the optimal time for prevention of dementia [15, 16]. In addition, residual confounding may still exist after adjustment for all covariates. Given these potential reasons, we selected midlife individuals younger than 65 years as the start of intervention for antihypertensive medication, and we specified the means of intervention at baseline instead of blood pressure change during follow-up time. Based on an emulated target trial using a cohort study design, our results suggested a causal relationship between early initiation of antihypertensive medication use and a reduction of incident dementia.

The different findings between observational studies and RCTs could be attributed to different features of their study designs for prevention of Alzheimer’s disease and related dementias, such as the eligibility criteria and sample composition, length of follow-up time, and treatment conditions [34]. The participants included in an RCT are highly selective, while there are few restrictions on observational studies. Also, due to the feasibility, the length of intervention in an RCT is usually shorter than 5 years, while it is common for observational studies to last for decades. In addition, treatment arms are randomly assigned to different groups, while the exposure in observational studies may be related to intrinsic characteristics of the groups of populations, which may remain after adjustment for covariates. All these differences do not only form the different definitions of the study designs, but also the different results and their interpretations. Taking the advantage of the large sample size and a long follow-up period of a cohort study, as well as methods of causal inference, we were able to estimate the effect of long-term intervention without the issue of limited feasibility. Admittedly, while the emulated target trial is a compromised way of conducting a real RCT, it does suggest a causal relationship, which answers a question of “Will the risk of dementia in late life be reduced if individuals with hypertension start to take antihypertensive medication from midlife” that can hardly be answered in an actual trial.

Different from the results of initiating antihypertensive medication on reducing the risk of incident dementia, our study did not show effectiveness for adhering to antihypertensive medication use on reduction of incident dementia. The different results between point and sustained effectiveness may suggest that the importance of antihypertensive medication in midlife, while use of antihypertensive medication in a later phase may not be as important as that in midlife. This is consistent with results from previous observational studies and RCTs. However, it did not necessarily mean there was no effectiveness of adherence to antihypertensive medication use on dementia, since the sample size may not be large enough to observe per-protocol effect, which can be reflected from the wide confidence intervals. This calls for future studies to estimate the effectiveness using large clinical datasets.

Our study has important clinical and public health significance. Using the causal inference methods, we were able to provide a causal relationship between antihypertensive medication use from midlife and prevention of incident dementia. This is consistent with our hypothesis that intervention on blood pressure reduction in midlife, instead of in late life, may be an effective way to prevent incident dementia. This also emphasizes the importance of midlife blood pressure in predicting a high risk of incident dementia in late life, as well as an early intervention on high blood pressure to prevent dementia. As is reflected in the effect sizes of these two emulated target trials, initiating antihypertensive medication use at an earlier stage of life could achieve much greater risk reduction on incident dementia. Also, with the use of antihypertensive medication, both cardiovascular and cognitive health will be improved among hypertensive patients, which may potentially increase the motivation of early initiation of antihypertensive medication and healthy lifestyle.