Twenty-four-hour versus clinic blood pressure levels as predictors of long-term cardiovascular and renal disease outcomes among African Americans

In Caucasian and Asian populations, evidence suggests that 24-h blood pressures (BP) are more predictive of long-term cardiovascular events than clinic BP. However, few long-term studies have evaluated the predictive value of 24-h BP phenotypes (24-h, daytime, nighttime) among African Americans (AA). The purpose of this study is to evaluate the added value of 24-h BP phenotypes compared to clinic BP in predicting the subsequent fatal and non-fatal cardiovascular/renal disease events in AA subjects. AA subjects (n = 270) were initially studied between 1994 and 2006 and standardized clinic BP measurements were obtained during screening procedures for a 3-day inpatient clinical study during which 24-h BP measurements were obtained. To assess the subsequent incidence of cardiovascular and renal disease events, follow-up information was obtained and confirmed by review of paper and electronic medical records between 2015 and 2017. During a mean follow-up of 14 ± 4 years, 50 subjects had one or more fatal or non-fatal cardiovascular/renal disease events. After adjustment for covariates, clinic systolic and diastolic BP were strongly associated with cardiovascular/renal disease events and all-cause mortality (p < 0.0001). Twenty-four-hour BP phenotypes conferred a small incremental advantage over clinic BP in predicting cardiovascular/renal events, which was limited to making a difference of one predicted event in 250–1,000 predictions depending on the 24-h BP phenotype. Nocturnal BP was no more predictive than the other 24-h BP phenotypes. In AA, 24-h BP monitoring provides limited added value as a predictor of cardiovascular/renal disease events. Larger studies are needed in AA to confirm these findings.


Results
Follow-up data on CV/renal disease and all-cause mortality were obtained for a total of 270 participants in whom 24-h BPs were measured. Those with at least one CV/renal disease event (morbidity and/or mortality) are referred to as cases and those without any of these events after at least a 10-year follow-up period are referred to as controls throughout this manuscript. Among the 50 cases, 72% were cardiac events (ischemic heart disease and/or congestive heart failure) and 16% were renal complications (end-stage renal disease, hemodialysis, and/ or renal transplant related to HTN). The remaining events included cerebrovascular events (52%), peripheral vascular disease (10%), and aortic dissection (6%). Forty-six percent of cases had more than one event (Supplemental Table 2). A total of 26 participants died from all causes (all-cause mortality), of which 12 were from CV/ Scientific RepoRtS | (2020) 10:11685 | https://doi.org/10.1038/s41598-020-68466-5 www.nature.com/scientificreports/ renal disease hence were part of the 50 cases mentioned above as well. Median length of follow-up was 12 years for the entire cohort (mean length of follow-up: 14.0 ± 4 years). Baseline characteristics (unadjusted) of all participants are shown in Table 1. Comparing the two groups, cases were older and had more men. Higher percentage of cases were hypertensive (90% vs. 47%) and were on anti-hypertensive medications (54% vs. 26%) compared to controls. After adjusting for age and sex, there were no differences in BMI, serum creatinine, glucose, insulin, lipids, aldosterone, and/or PRA levels. Table 2 shows BP levels of all participants during the baseline clinic visit. All BP phenotypes (clinic BP, 24-h BP, daytime BP, and nighttime BP) were higher in those who were defined as cases compared to controls. The correlations (r) of clinic BPs with 24-h BPs (both SBP and DBP) were 0.75 (p < 0.0001). Both clinic and 24-h SBP phenotypes were predictive of all-cause mortality along with daytime and nighttime DBP using logistic regression analyses. Using cox proportional hazard model, clinic and nighttime systolic and diastolic BPs were able to Table 1. Comparison of baseline characteristics of participants with and without cardiovascular/renal disease outcomes (unadjusted). All values are expressed as mean ± SD unless otherwise specified. Conversion to SI units (multiplication factor): total cholesterol, LDL-C, and HDL-C: 0.0259 (mmol/L), HDL-C, triglycerides: 0.0113 (mmol/L), glucose: 0.05555 (mmol/L), insulin: 6.945 (pmol/L), creatinine: 88.4 (µmol/L), PRA (pg/ mL): 0.0237 (pmol/L), aldosterone: 27.74 (pmol/L). SD standard deviation, BMI body mass index, HDL-C high-density lipoprotein-cholesterol, LDL-C low-density lipoprotein cholesterol, PRA plasma renin activity. *p < 0.05; **p ≤ 0.01; ***p ≤ 0.001; ****p ≤ 0.0001. a Expressed as percentages.  Fig. 2A-F). To further analyze how much more value 24-h BP phenotypes add to clinic BPs in predicting one CV/renal disease event, we calculated the average predicted probabilities of having an event for each of the BP phenotypes (Table 3, Supplemental Fig. 1A-B). In all participants, clinic BP measurements provided a slightly higher average probability of predicting an event than using 24-h BP phenotype (equivalent to 1 more event in 333 predictions), i.e. using clinic SBP (or DBP) would predict one more event than using 24-h SBP (or DBPs in 333 predictions). Using daytime SBP (or DBP) gives the same predictive probabilities as using the clinic SBP (or DBP). Using nighttime SBP we would predict 1 more event than using clinic SBP in 1,000 measurements, while nighttime DBP will predict 1 fewer event than using clinic DBP in 333 measurements. The findings did not differ when only participants not on anti-hypertensive medications were analyzed (see supplemental data). The results also did not differ when we analyzed the data after adjusting the BP levels for anti-hypertensive medications (also shown in supplemental data).
Nocturnal BPs. Overall, African American subjects in our study showed less than 10% nocturnal dip. The magnitude of the nighttime BP dip did not differ between participants who experienced a CV/renal disease event and those who did not (Fig. 1). Day-nighttime differences were not predictive of composite CV/renal disease events or all-cause mortality . The odds ratio for predicting a CVD/renal disease event based on nighttime BP was no more robust than predictions based on other 24-h BP phenotypes (Supplemental Tables 3 and 5).

Discussion
In this study of an African American cohort, we found that both outpatient clinic BPs and 24-h BP phenotypes are predictive of long-term composite CV/renal disease morbidity and mortality. Moreover, measurement of 24-h BP phenotypes provided limited incremental value as a predictor of CV/renal disease risk among African Americans, by differing in prediction by one less or more event in ~ 250 (range 250-1,000 depending on the 24-h BP phenotype) predictions, compared to clinic BP values. Even though several similar studies comparing clinic and 24-h BPs have been reported in Caucasian and Asian populations 24,25 , this is the first study in African Americans that directly compared the predictive effects of outpatient BP levels versus 24-h BP phenotypes in predicting CV/renal disease events 32 .
There has been no question of superiority of ABPM levels in improving the accuracy, minimizing the error, and standardization of measurements due to usage of automated devices in both African Americans and Caucasians 33,34 . ABPM is also essential to the diagnosis of various HTN phenotypes including white-coat, masked, and/or nocturnal in addition to BP variability 4,9,14,16 . With ABPM, it is also possible to assess the efficacy of anti-hypertensive medications throughout the day and night rather than at a single time-point with measurements of BPs in the office 35,36 .  www.nature.com/scientificreports/ Starting in early 1980s, several thousands of patients with ABPM have been followed for varying durations of time resulting in the conclusion that levels of ABPM phenotypes (among them nocturnal HTN) are better predictors of subclinical organ damage and CVD morbidity and mortality than the corresponding clinic BP   40 . Some recent studies have questioned this heightened risk with white-coat HTN and suggest that it may not be higher than their normotensive control subjects after adjustment of other risk factors [41][42][43] . In addition, masked HTN has been associated with increased CVD risk factors [43][44][45] . Most of these studies have been conducted in European and Asian populations 32 . In addition, ABPM levels are also valuable in evaluating BP variability which has been associated with target-organ damage 46,47 . Few studies have evaluated the effect of ABPM on long-term CV/renal disease morbidity and mortality in African American individuals when compared to clinic BP values 32 . Most notable of this is Jackson Heart Study (JHS), which has shown that various 24-h BP phenotypes such as lack of nocturnal dipping, BP variability, and masked HTN were negatively associated with intermediate target-organ damage markers such as glomerular filtration rate decline (chronic kidney disease), increased left ventricular mass index, and increased carotid artery intimal-media thickness when controlled for clinic BPs indicating superiority of ABPM in predicting these intermediate outcomes [48][49][50][51][52][53] . In a recent study, JHS reported increased predictability of day time and night time BPs in prediction of CV events and all-cause mortality, however, they did not report on the incremental advantage bestowed by cumbersome 24 h BP measurements and how many events would be predicted by utilizing 24 h BP phenotypes 54 .
In this study, we compare the predictability of composite CV/renal disease outcomes and all-cause mortality by clinic BPs versus 24-h BP phenotypes among African Americans. While we found that all BP phenotypes were predictive of long-term outcomes, the incremental value of 24-h BP phenotypes over clinic BP measurements was limited and the average predictive difference was one less or more predicted CV/renal disease event in 250 individuals (range 250-1,000) compared to clinic BPs. Our findings are in contrast to findings from several studies that indicated that 24-h BPs indeed add a significant value over clinic BPs and save lives. However, almost all of these studies that have evaluated morbidity and mortality data were conducted in Caucasian and Asian subjects. Among studies in African American subjects, the findings are from evaluation of intermediate target-organ damage parameters rather than actual CV/renal disease events and in those that reported this data-the incremental advantage was not specified. In a recent publication based on JHS participants, a total of 165 events (including CVD events [n = 80] and all-cause mortality [n = 85]) were identified after a median 10.8-year follow-up among individuals who underwent ABPM. Five-year predicted probability of a CVD or an all-cause mortality event (combined) in this study was associated with increasing levels of clinic BPs, however, no comparison was made with ABPM phenotypes in this study 55 . Another study on masked HTN showed that CVD risk for masked HTN was not heightened compared clinic BPs in African Americans; C-statistic for CVD risk with masked HTN and clinic BPs being 0.681 and 0.703 respectively (higher the C-statistic, better the prediction model) 56 . Most recent study from JHS showed that only nighttime SBP was associated with modest increase in all-cause mortality 54 . In addition, hazard ratios for cardiovascular events was increased only with large BPs differences (10-16 mmHg) between 24 h phenotypes and clinic phenotypes 54 . These data from JHS and findings from our study suggest 24-h BP phenotypes may not be much superior to clinic BPs in predicting CV/renal disease events among African Americans.
An explanation for these findings could be due to higher 24-h BP levels seen among African Americans compared to Caucasians and Asians, particularly lack of day-night dip. Levels of ABPM phenotypes (24-h, daytime, and nighttime) are typically lower than clinic BP values with a clinic BP value of 140/90 mmHg approximately corresponding to a 24-h BP of 130/80 mmHg, daytime BP of 135/85 mmHg, and nighttime BP of 120/70 mmHg, values determined based on several studies 33 . However, the JHS found that outcome-derived ABPM thresholds for African Americans were higher than those from published recommendations for European, Asian, and South American populations with a clinic SBP ≥ 140 mmHg corresponding to 24-h, daytime, and nighttime SBP of 138 mmHg, 134 mm Hg, and 129 mmHg respectively 55 . We found similarly attenuated differences between clinic and 24-h BP phenotypes in our study population. As an example, albeit the measurements were separated Table 3. Average predictive probability of CV/renal event and comparison of predictions of various BP phenotypes. CV cardiovascular, BP blood pressure, SBP systolic blood pressure, DBP diastolic blood pressure. a One more prediction when compared to clinic BP. b One less prediction when compared to clinic BP.

BP variables
Average predicted probability in all-participant cohort  Table 4). It is plausible that this smaller difference between clinic and 24-h BP phenotypes negated whatever advantage was conferred by 24-h BPs in other ethnic groups.
In several racial/ethnic populations, nighttime BP has been shown to have the strongest prognostic value for CVD, compared with 24-h, daytime, and clinic BPs 22,57,58 . It is well-known that, compared with other ethnic groups, African Americans have a blunted nighttime BP dip 27,28 . While there are no outcome studies associating nocturnal HTN or non-dipping patterns with CV/renal disease morbidity and mortality, nocturnal BPs were shown to be associated with some intermediate markers of target organ damage among African Americans 49 . In the current study, we found that the nocturnal BP dip did not differ between cases and controls, and there was no evidence of superiority of nocturnal BP compared to other 24-h BP phenotypes and only provided modest improvement over clinic BPs as a predictor of CV/renal disease risk.
There are several strengths to this study. All CV/renal disease events were verified and adjudicated by reviewing patient medical records by board-certified physicians. In addition, medical records of all individuals who were determined to have not had an event for 10 years were also reviewed prior to designating them as controls unlike many of the other studies that have only verified positive events 32 . Despite its strengths, we acknowledge our study's limitations. Our sample size is moderate. We do not have reliable information on history of smoking, alcohol use, or physical activity. Our definition of daytime and nighttime are based on fixed-clock times, which by including transitional times-the hours during which some patients may be asleep, whereas others are awake-may represent a source of mis-classification. However, studies have shown that the day and night durations defined by wide fixed-clock intervals (6:00 am to 10:00 pm for the day and 10:00 pm to 6:00 am for the night), narrow fixed-clock intervals (10:00 am to 8:00 pm for the day and midnight to 6:00 am for the night) or according to patient's diary does not affect predictability of in-study outcomes 59 . Other limitations include measurement of 24-h BP levels were conducted in an in-patient facility rather than during their daily activities. Based on our BP levels which are comparable to what was reported to JHS, these limitations did not adversely affect the BP measurements.
Based on extensive reviews of the available data, European and American guidelines recommend that ABPM be considered an "adjunct"" to the "gold standard" clinic measurement 7,60 . While the value of ABPM in diagnosing and monitoring BP is undisputable, its real utility is in being a better predictor of CV/renal disease compared to clinic BPs. However, we found that 24-h BP phenotypes are only marginally better in predicting CVD events compared to clinic BPs among African Americans with 24-h BPs predicting one less or more CV/renal event in 250 individuals (range 250-1,000). It is plausible that this smaller difference compared to Asian and European populations is due to less significant differences between the levels of clinic BPs and 24-h BP phenotypes in African Americans. This raises questions of cost-effectiveness of ABPM and added burden of 24-h BP measurement among African Americans. In conclusion, in middle aged African Americans, our results suggest that a carefully measured clinic BP is nearly as effective as 24-h BP measurement in predicting CV/renal disease events and ABPM is of limited incremental value. Larger studies in African Americans are needed to confirm these findings.

Data availability
All data will be provided in excel spreadsheet without restriction upon request.