Blood pressure and cognitive function across the eighth decade: a prospective study of the Lothian Birth Cohort of 1936

Objectives We investigated the associations among blood pressure and cognitive functions across the eighth decade, while accounting for antihypertensive medication and lifetime stability in cognitive function. Design Prospective cohort study. Setting This study used data from the Lothian Birth Cohort 1936 (LBC1936) study, which recruited participants living in the Lothian region of Scotland when aged 70 years, most of whom had completed an intelligence test at age 11 years. Participants 1091 members of the LBC1936 with assessments of cognitive ability in childhood and older adulthood, and blood pressure measurements in older adulthood. Primary and secondary outcome measures Participants were followed up at ages 70, 73, 76 and 79, and latent growth curve models and linear mixed models were used to analyse both cognitive functions and blood pressure as primary outcomes. Results Blood pressure followed a quadratic trajectory in the eighth decade: on average blood pressure rose in the first waves and subsequently fell. Intercepts and trajectories were not associated between blood pressure and cognitive functions. Women with higher early-life cognitive function generally had lower blood pressure during the eighth decade. Being prescribed antihypertensive medication was associated with lower blood pressure, but not with better cognitive function. Conclusions Our findings indicate that women with higher early-life cognitive function had lower later-life blood pressure. However, we did not find support for the hypothesis that rises in blood pressure and worse cognitive decline are associated with one another in the eighth decade.

pressure score, as well as anti-hypertensive medication data, which were used to 63 adjust blood pressure measurements.

64
-This study had comprehensive tests of cognitive ability measured in both childhood 65 and old age which allowed us to investigate whether childhood and old age cognitive 66 ability are distinctly related to blood pressure.

67
-Latent growth curve modelling allowed us to evaluate whether changes in either  93 These latter findings are part of a field known as cognitive epidemiology, which has found 94 that higher cognitive function in early life is associated with lower risk of a number of physical 95 and mental ailments later in life. [12][13][14][15][16] 96 Men are more likely to develop cardiovascular conditions than women, 17 a reason 97 why men have been the subject of more intervention studies than women. 18  Sex and education were both included as covariates in all models. Education was 171 recorded as the total number of years spent in formal education. Age 11 cognitive function 172 was also included in all models, as was the interaction between sex and age 11 function.

173
Several behavioural and diagnosed health variable were included. Smoking status 174 was dummy coded with two variables, current smokers and ex-smokers, with non-smokers 175 as the reference group. Participants were genotyped for the presence of an APOE ε4 allele.

176
Cardiovascular disease and stroke history were both recorded for each wave; each was 177 coded as a binary variable, with a 1 indicating that the individual had a history of the disease.

178
Testing dates varied slightly for every individual at every wave, so the exact age of each 179 participant at testing was also recorded. analyse multiple latent and measured variables. 39 For each variable modelled longitudinally,

185
LGCMs model level (i.e. an intercept) and slope (i.e. a trajectory of change) parameters. We

198
We modelled cognitive function using a slightly different hierarchical 'factor of 199 curves' model, previously established with these data. 30 For each cognitive test we modelled

236
The results will be disseminated to participants via a quarterly newsletter sent to LBC1936 237 participants.

240
Modelling blood pressure

241
The best fit model of BP was a good fit for these data (χ² = 2052.76, df = 259, P <

245
Blood pressure scores for each wave were derived from the latent variables and 246 illustrated in Figure 1. In both hypertensive and normotensive individuals, blood pressure 247 scores were similar at each age except for a notable but unsurprising difference in 248 magnitude: individuals with hypertension had much higher blood pressure than those who 249 did not have hypertension. Both groups show a quadratic effect, blood pressure rose 250 between ages 70 and 76, and then declined between age 76 and 79.  will still be strongly associated.

270
Third, a significant regression path was found from the interaction of age 11 cognitive 271 function and sex to eighth decade blood pressure level (β = -0.136, SE = 0.050, P = 0.047).

272
This interaction association indicates that women with higher cognitive function at age 11

278
Incorporating control variables into the LGCM (Model 1B, supplementary table 3) 279 reduced the regression weight from the sex and age 11 cognitive function interaction to 280 blood pressure level, so that it was no longer significant (β = -0.106, SE = 0.058, P = 0.215).

281
The variable that most likely caused this change in the model was smoking behaviour; 282 specifically, being a current smoker at age 70 had an association with eighth decade blood In an effort to understand why we found no associations between cognitive function 287 and blood pressure in the eighth decade, we examined the role of medication in post-hoc 288 analyses. Most individuals diagnosed with hypertension in the LBC1936 were prescribed 289 medication for the condition. Following from Figure 1, though individuals with hypertension 290 would have had higher blood pressure were they not being treated, they are being treated,

291
and their blood pressure appears to generally decline over the eighth decade.

292
We fit linear mixed effect models to explore the associations of diagnosis and 293 medication with blood pressure. Using latent variables for blood pressure, we generated 294 blood pressure scores at each wave, but, by contrast with the previous analyses, the scores 295 were based on the raw, medication-unadjusted systolic and diastolic blood pressure 296 measurements. Using these scores, we first fit a mixed model predicting blood pressure 297 score from a 2-way model that included the wave of the study, whether an individual had a 298 hypertension diagnosis at that wave, and the interaction of the two (Model 2A, Table 2). We 299 found a significant main effect of wave on blood pressure score (β = -0.048, SE = 0.014, P < 300 0.001) and interaction between wave and hypertension diagnosis (β = -0.111, SE = 0.028, P 301 < 0.001), suggesting that individuals who are given a hypertension diagnosis have 302 decreasing blood pressure over time.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59 1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59   Only study completers, those participants who were present at every wave of the study, were included in this plot. All study participants were used in the statistical analyses. Blood pressure score was derived from the latent variable that represent blood pressure in individual waves. Blood pressure score is a standardized, unitless measure of overall blood pressure magnitude; A score of 0 indicates the individual is in line with average blood pressure, and 1 would indicate that an individual was 1SD higher blood pressure at that time. The upper green line illustrates individuals diagnosed with hypertension at any wave (n = 337) and the lower red line illustrates normotensive individuals (n = 204). The shaded areas represent 95% confidence regions.

Introduction
Background/rationale 2 Explain the scientific background and rationale for the investigation being reported 5, 6 Objectives 3 State specific objectives, including any prespecified hypotheses 6

Study design 4
Present key elements of study design early in the paper 6, 7 Setting 5 Describe the setting, locations, and relevant dates, including periods of recruitment, exposure, follow-up, and data collection 6, 7 (a) Give the eligibility criteria, and the sources and methods of selection of participants. Describe methods of follow-up 6, 7 Participants 6 (b) For matched studies, give matching criteria and number of exposed and unexposed NA

116
Some cognitive functions steadily decline in mean levels in older participants, while 117 hypertension is a condition that becomes increasingly common with age and is related to 118 cardiovascular health and cognitive impairment. In light of this, we tested two hypotheses 119 regarding the relationships between cognitive functions and blood pressure in the present 120 study. First, we hypothesized that the association between higher cognitive function at age 121 11 years and lower blood pressure in the eighth decade of life is stronger in women than it is 122 in men. 11 Second, we tested the hypothesis that cognitive function and blood pressure will 123 be reciprocally associated with each other across the eighth decade. We were able to test 124 these hypotheses using multi-wave data from the Lothian Birth Cohort 1936 (LBC1936), a 125 narrow-age cohort of over 1000 community-dwelling people, and longitudinal modelling of     LGCMs to account for measurement error, 45

260
The results will be disseminated to participants via a quarterly newsletter sent to LBC1936 261 participants.

264
Modelling adjusted blood pressure

265
The best fit model of BP adjusted for medication was a good fit for these data (χ² =

270
Blood pressure scores for each wave were derived from the latent variables and 271 illustrated in Figure 1. In both hypertensive and normotensive individuals, blood pressure 272 scores were similar at each age except for a notable but unsurprising difference in  will still be strongly associated.

295
Third, a significant regression path was found from the interaction of age 11 cognitive 296 function and sex to eighth decade blood pressure level (β = -0.136, SE = 0.050, P = 0.047).

297
This interaction association indicates that women with higher cognitive function at age 11 298 have lower blood pressure in later life, and women with lower cognitive function have higher 299 blood pressure. The opposite was true for men: males with higher cognitive function at age 300 11 had higher blood pressure and vice versa (Figure 3). No additional associations were 301 found between any specific cognitive domains and blood pressure variables. All these tested 302 associations are visualized in Figure 2.

303
Incorporating control variables into the LGCM (Model 1B, supplementary table 3) 304 reduced the regression weight from the sex and age 11 cognitive function interaction to 305 blood pressure level, so that it was no longer significant (β = -0.106, SE = 0.058, P = 0.215).

306
The variable that most likely caused this change in the model was smoking behaviour; specifically, being a current smoker at age 70 had an association with eighth decade blood 308 pressure (β = -0.104, SE = 0.040, post-hoc uncorrected P = 0.009).

309
Unadjusted blood pressure and cognitive function 310 We also attempted to fit bivariate LGCMs of unadjusted blood pressure and cognitive 311 function. These models were intended to be the same as the models discussed above, but 312 without the blood pressure variables being adjusted when individuals were recorded as 313 taking anti-hypertensive medication. These models would not converge, and we were unable 314 to progress further with this analysis. In order to investigate the influence of medication 315 further, we carried out cross-sectional and additional longitudinal analyses.

Cross-sectional correlations between cognitive functions and blood pressure scores across
318 waves

319
Given the quadratic trajectory of blood pressure over time in these data, and earlier 320 evidence suggesting that the association between cognitive function and blood pressure 321 changes depending on age, 49 we examined correlations between cognitive function, both 322 domains and general ability, and blood pressure scores, both with and without adjustment 323 for anti-hypertensive medication. We carried out these correlations within wave, that is, we 324 looked at correlations between age 70 cognitive function and age 70 blood pressure, age 73 325 cognitive function and age 73, blood pressure, and so on for ages 76 and 79.

326
The results of these analyses are shown in Figure 4 and supplementary table 4.

327
Correlations were similar at all time points, although all correlations were also small; no 95% 328 confidence intervals did not overlap with zero. In an effort to understand why we found no associations between cognitive function 332 and blood pressure in the eighth decade, we examined the role of medication in post-hoc 333 analyses. The majority of individuals diagnosed with hypertension in the LBC1936 were 334 prescribed medication for the condition. Following from Figure 1, though individuals with 335 hypertension would have had higher blood pressure were they not being treated, they are 336 being treated, and their blood pressure appears to generally decline over the eighth decade.

337
We fit linear mixed effect models to explore the associations of diagnosis and 338 medication with blood pressure. Using latent variables for blood pressure, we generated 339 blood pressure scores at each wave, but, by contrast with the previous analyses, the scores 340 were based on the raw, medication-unadjusted systolic and diastolic blood pressure 341 measurements. Using these scores, we first fit a mixed model predicting blood pressure 342 score from a 2-way model that included the wave of the study, whether an individual had a 343 hypertension diagnosis at that wave, and the interaction of the two (Model 2A, Table 2). We

372
In the second model of this set of models (Model 3B,

385
Our results show that sex and cognitive function from early-life interact to predict 386 blood pressure level during the eighth decade. Women with higher cognitive function are 387 less likely than higher cognitively functioning men to have higher blood pressure, but the 388 opposite is also true: lower cognitive function women are more likely to have higher blood 389 pressure than similarly functioning men (Figure 3). This finding provides mixed support for

395
We did not find compelling evidence to support our second hypothesis. There were 396 no associations between high blood pressure and cognitive function, either in level or

Introduction
Background/rationale 2 Explain the scientific background and rationale for the investigation being reported 5, 6 Objectives 3 State specific objectives, including any prespecified hypotheses 6

Study design 4
Present key elements of study design early in the paper 6, 7 Setting 5 Describe the setting, locations, and relevant dates, including periods of recruitment, exposure, follow-up, and data Describe any efforts to address potential sources of bias 10 Study size 10 Explain how the study size was arrived at 6, 7 Quantitative variables 11 Explain how quantitative variables were handled in the analyses. If applicable, describe which groupings were chosen and why 9, 10 (a) Describe all statistical methods, including those used to control for confounding 9, 10 (b) Describe any methods used to examine subgroups and interactions 9, 10 (c) Explain how missing data were addressed 10 (d) If applicable, explain how loss to follow-up was addressed 10 Statistical methods 12 (e) Describe any sensitivity analyses NA  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46