Blood pressure trajectories in young adulthood and midlife steatotic liver disease

Abstract

Background:

Although high blood pressure (BP) is a risk factor for steatotic liver disease (SLD), whether longitudinal BP trajectories starting in young adulthood are associated with midlife SLD is unclear.

Methods:

We used data from a U.S.-based prospective cohort, the Coronary Artery Risk Development in Young Adults (CARDIA) study, which enrolled 5115 Black and White adults aged 18 to 30 years in 1985-1986 (Year 0, Y0). Diastolic and systolic BP, hypertension trajectories were evaluated based on up to eight BP measures over 25-year follow-up. SLD at Y25 was measured using non-contrast abdominal computed tomography scans.

Results:

Among 2817 participants included in analytic sample, 622 SLD cases occurred. Compared with low-stable diastolic BP group, other trajectories showed higher SLD prevalence (22.0%-37.5% vs. 12.8%), with multivariable-adjusted relative risks of 1.47 (95% CI: 1.17-1.84) for moderate-stable, 1.77 (1.33-2.35) for moderate-increasing (remaining below the threshold for hypertension), 1.57 (1.21-2.04) for elevated-stable, and 2.02 (1.41-2.88) for elevated-increasing trajectory. Similar positive associations were observed for systolic BP trajectories. Compared to those persistently normotensive, participants who developed hypertension had significantly higher SLD risks: 1.62-fold for late-onset, and 1.75-fold for both mid-adulthood and early-onset. Consistently, a faster rate of BP increase was linked to higher SLD risk.

Conclusion:

Long-term BP and hypertension patterns in young adulthood were independently associated with increased midlife SLD risk, with rapidly increasing BP exhibiting the greatest risk. Long-term BP changes may assist in more accurate identification of individuals at higher risk of SLD.

Graphical Abstract

Introduction

Steatotic liver disease (SLD), affecting over 30% of adults in the U.S.,¹ is now the ninth leading cause of death,² and results in $103 billion in annual direct medical costs.³ As treatment and prevention for viral hepatitis improve, non-viral causes of SLD, such as metabolic dysfunction-associated, are becoming the leading cause of liver-related morbidities and an indicator for liver transplantation.^4,5

Hypertension affects approximately half of U.S. adults,⁶ with up to 50% of those with hypertension also experiencing SLD.⁷ Evidence from longitudinal studies links elevated blood pressure (BP) levels to the development and progression of SLD.^8–11 Mendelian randomization (MR) studies also supports the causal effect of hypertension on SLD,¹² independent of genetically predicted body mass index (BMI). The shared pathophysiological mechanisms—such as visceral obesity, insulin resistance (IR), and chronic inflammation—may explain this overlap.¹³ However, although BP levels increase with aging may begin in young adulthood,¹⁴ most previous observational studies have measured BP only once in mid- or late life.^8,9 Consequently, liver damage associated with prolonged BP elevation may start well before midlife. Trajectory analysis is an approach that captures clusters of individuals with similar change patterns in the measure of interest over time, providing a more accurate representation of long-term exposure.¹⁵ Previous studies have shown that BP trajectories over time are heterogeneous, with some individuals maintaining stable BP at varying levels while others experience faster increases.^16–18 Greater increases from youth have been linked to higher lifetime risk of cardiovascular disease,¹⁶ poor brain health,¹⁸ and renal damage.¹⁷ Investigating long-term BP and hypertension trajectory patterns starting in young adulthood in relation to midlife SLD risk may offer important insights into identifying subgroups at higher risk of later SLD and potential windows of opportunity for early intervention.

Therefore, we aimed to identify individuals’ long-term trajectories of diastolic BP, systolic BP, and hypertension starting in young adulthood, and prospectively evaluate their associations with midlife SLD risk, using data from the Coronary Artery Risk Development in Young Adults (CARDIA) study. We hypothesized that, in comparison to a trajectory in which individuals maintain ideal BP levels or remain normotensive throughout young adulthood, those experiencing higher BP levels, faster rates of BP increase, or earlier onset of hypertension would have a greater risk of SLD in midlife. In secondary analysis, we quantified the association between annual BP increase rate and SLD risk.

Methods

Study Design and Population

The CARDIA study is an ongoing prospective cohort study that has been described in detail elsewhere.¹⁹ Briefly, 5115 self-identified Black and White men and women aged 18 to 30 years were recruited from four U.S. field centers (Birmingham, Alabama; Chicago, Illinois; Minneapolis, Minnesota; and Oakland, California) in 1985-1986 (Year 0, Y0), and have been followed up through in-person examinations at regular intervals for more than 35 years, with high retention rates.²⁰ All study protocols were approved by institutional review boards at each field center, and all participants provided written informed consent at each examination.

BP and hypertension trajectories were modeled among 4680 CARDIA participants with at least three BP measures from Y0 to Y25 (80% had ≥6 measurements). For analyses of BP trajectories and SLD, we excluded participants with possible or unknown SLD at Y0 (identified using the fatty liver index [FLI] ≥60,²¹ calculated based on BMI, waist circumference, gamma-glutamyl transferase, and triglyceride levels; n=395; see Supplemental Methods), self-reported liver disease at Y0, missing BP trajectories data, missing liver attenuation (LA) measured, or missing important covariates. The final analytic sample included 2817 participants (Figure 1). Characteristics and BP trajectories of participants with and without LA measurements are shown in Table S1–S2.

Figure 1.

Flow chart for selecting the Coronary Artery Risk Development in Young Adults study participants for analysis Data was presented as the number (the percentage of the original cohort).

BP measurements

BP was measured by trained technicians at eight examinations over 25 years using standardized protocols, with a Hawksley random-zero sphygmomanometer (W A Baum Company, Copiague, NY) used from Y0 to Y15 and a digital BP monitor (Omron HEM907XL; Online Fitness, Santa Monica, CA) used from Y20 to Y25.¹⁶ The Hawksley devices were inspected weekly and regularly calibrated, while the Omron devices were checked regularly. The digital monitor readings have been calibrated to the sphygmomanometer measures to minimize machine bias.²² The right arm BP was taken as average of the second and third of 3-time measurements at a 1-minute interval after the participant had been sitting quietly for 5 mins in a still room. Hypertension was defined as systolic BP ≥140 mmHg, and/or diastolic BP ≥90 mmHg, and/or current use of antihypertensive medications.²³ Use of antihypertensive medications was collected through self-reported standardized medication inventory questionnaires at each examination.

SLD measurement

Abdomen non-contrast multidetector computed tomography (CT) scans were performed to measure LA at Y25.²⁴ The LA was measured in Hounsfield units (HU) as average of nine measurements in liver right lobe on three CT slices using circular regions of interest of 2.6 cm².²⁴ The intraclass correlation coefficient for LA between different trained readers was high (0.975, random sample n=156). Any SLD at Y25 was defined as LA of less than 51 HU, approximating a liver-to-spleen ratio less than 1.0, which indicates mild or greater steatosis is present (83.3% specificity, 75.0% sensitivity).^25,26 The SLD was further categorized into metabolic dysfunction-associated steatotic liver disease (MASLD) and other subtypes (metabolic dysfunction and alcohol-associated liver disease [MetALD], alcohol-associated liver disease [ALD], and other causes of SLD) (see Supplemental Methods).²⁷ Moderate-to-severe SLD was defined as SLD cases with a LA less than or equal to 40 HU (>30% steatosis),²⁵ with reported specificity of 100% and sensitivity of 82%.^26,28

Covariates

In CARDIA, data on age, sex, self-reported race, field center, education, family income, smoking status, alcohol intake, physical activity, dietary intake, BMI, waist circumference, total cholesterol, fasting glucose, fasting insulin, C-reactive protein (CRP), creatinine, albumin, self-reported cirrhosis and other liver diseases were collected. Homeostasis model assessment of IR (HOMA-IR),²⁹ estimated glomerular filtration rate, and urine albumin-creatinine ratio³⁰ were calculated. See Supplemental Methods.

Statistical analysis

The primary exposures were long-term patterns in BP and hypertension over 25 years, identified by the group-based trajectory modeling (Supplemental Methods).¹⁶ This method allows handling of missing BP data for each individual. Five 25-year trajectories were identified for both diastolic and systolic BP: low-stable (maintained low levels), moderate-stable (starting moderate levels and remaining stable), moderate-increasing (starting moderate levels then gradually increasing), elevated-stable (starting high levels and remaining stable), and elevated-increasing (starting high levels then gradually increasing). For hypertension, four trajectories were identified and labelled as: mostly normotensive (maintaining normal BP throughout 25 years), early-onset (having hypertension at Y0 or developing hypertension before age 35), mid-adulthood onset (developing hypertension between ages 35-45), and late-onset (developing hypertension after age 45). As a secondary exposure, the annual BP increase rate was derived as slope of a linear regression of BP on age over 25 years.

Characteristics of participants by BP and hypertension trajectories are presented as mean (SD) for normal continuous variables, median (IQR) for skewed continuous variables, and proportions for categorical variables. Modified Poisson regression with robust standard errors was used to assess relative risks (RRs) and 95% confidence intervals (CIs) for the associations between trajectories and SLD risk. We conducted three multivariable-adjusted analyses for, separately, diastolic BP, systolic BP, and hypertension trajectories: Model 1 adjusted for baseline age, sex, self-reported race, and field center; Model 2 further adjusted for baseline education, family income, smoking status, alcohol intake, physical activity, Healthy Eating Index-2015, BMI, fasting glucose, total cholesterol level; and Model 3 further adjusted for antihypertensive medications use over 25 years. In secondary analysis, we evaluated the association of annual BP increase rate with SLD risk. We also further examined the associations of BP and hypertension trajectories with MASLD, other SLD subtypes, and moderate-to-severe SLD.

We modeled sex-specific trajectories to explore potential sexual dimorphism in the associations of BP trajectories and SLD. To further assess the robustness of our findings, we conducted several sensitivity analyses and stratified analyses by self-reported race, healthy lifestyle score, BMI, and genetic susceptibility of SLD³¹ (Supplemental Methods and Table S3).

All analyses were performed in R 4.3.1 (R Foundation for Statistical Computing, Vienna, Austria) and SAS version 9.4 (SAS Institute Inc). A two-sided P <0.05 was considered statistically significant.

Results

Five common long-term diastolic BP trajectories were identified (Figure 2A): 1) low-stable: 992 (21.2%) individuals (with mean values from 60.7 mmHg at Y0 to 63.8 mmHg at Y25); 2) moderate-stable: 2113 (45.1%) individuals (67.5 to 74.1 mmHg); 3) moderate-increasing: 450 (9.6%) participants (68.0 to 90.1 mmHg, remaining below the threshold for clinical hypertension); 4) elevated-stable: 928 (19.8%) individuals (76.7 to 77.8 mmHg); and 5) elevated-increasing: 197 (4.2%) individuals (82.9 to 91.3 mmHg). Similar patterns were observed for systolic BP level (Figure 2B), with participants in the moderate-increasing and elevated-increasing trajectories experiencing the largest increases of 36.0 mmHg (111.8 to 147.8 mmHg) and 18.3 mmHg (127.2 to 145.5 mmHg) over 25 years, respectively. For the presence or absence of hypertension at each examination, four trajectories were identified: mostly normotensive (65.3%, n=3054), late-onset hypertension (19.3%, n=903), mid-adulthood onset hypertension (11.8%, n=552), and early-onset hypertension (3.7%, n=171) (Figure 2C). Detailed BP measures and hypertension prevalence by trajectories are presented in Table S4.

Figure 2.

Trajectories of estimated blood pressure and hypertension in the Coronary Artery Risk Development in Young Adults Study

Of these, 2817 eligible participants were used for main analyses, with a mean (SD) age of 25.0 (3.6) years at Y0. Participants consisted of 43.1% men and 56.9% women, with 54.0% White and 46.0% Black adults. Participants’ characteristics for BP and hypertension trajectories are presented in Table S5–S6 and Table 1. Compared with participants in the mostly normotensive trajectory, those with earlier onset hypertension were more likely to be Black, have lower educational attainment and family income, engage in less physical activity, and have poorer dietary quality. Additionally, they exhibited higher BMI and waist circumference levels at Y0, as well as greater weight gain and glucose increase over 25 years.

Table 1.

Characteristics of participants by 25-year hypertension trajectories: The Coronary Artery Risk Development in Young Adults Study

Characteristic	Total	Hypertension trajectories
		Mostly normotensive	Late-onset hypertension	Mid-adulthood onset hypertension	Early-onset hypertension
N	2817	1807	619	315	76
Age, years	25.0 (3.6)	24.9 (3.6)	26.0 (3.3)	23.9 (3.6)	24.6 (3.5)
Men, %	43.1	42.2	46.5	40.3	47.4
White, %	54.0	63.4	44.9	23.8	28.9
Field center, %
Birmingham	22.9	19.2	26.8	35.2	26.3
Chicago	22.6	24.2	19.4	21.0	17.1
Minneapolis	26.8	29.2	25.4	16.5	23.7
Oakland	27.8	27.4	28.4	27.3	32.9
Education, %
High school or less	34.7	31.8	37.5	43.5	43.4
Some college	33.7	31.8	35.5	40.6	36.8
College graduate or higher	31.6	36.4	27.0	15.9	19.7
Family income at Y5,a %
Less than $25,000	32.9	29.8	34.6	44.8	44.7
$25,000 to $50,000	35.9	36.7	37.8	27.9	34.2
$50,000 and greater	24.0	26.8	19.9	18.4	13.2
Missing	7.2	6.7	7.8	8.9	7.9
Smoking status, %
Never	60.3	61.5	54.0	64.8	64.5
Former	13.5	14.2	14.4	8.9	9.2
Current	26.2	24.3	31.7	26.3	26.3
Alcohol intake, median (IQR), drink/wk	2.1 (0-6.2)	2.1 (0-6.2)	2.4 (0-7.2)	1.0 (0-4.4)	2.2 (0-7.4)
Never/Former	38.1	37.6	36.2	44.1	39.5
Current: ≤ 7(W)/14(M)	46.8	46.9	48.1	45.4	40.8
Current: > 7(W)/14(M)	15.1	15.6	15.7	10.5	19.7
Physical activity, exercise units	424 (302)	438 (294)	418 (321)	371 (297)	364 (308)
Heathy eating index-2015	58.6 (10.2)	59.5 (10.2)	57.5 (10.0)	56.7 (10.3)	55.3 (9.8)
BMI, kg/m2	23.6 (3.5)	23.1 (3.2)	24.5 (3.7)	24.7 (4.0)	25.0 (4.5)
Waist circumference, cm	75.8 (8.7)	74.6 (8.4)	77.8 (8.6)	77.4 (9.1)	79.2 (9.2)
Fasting glucose, mg/dL	81.7 (10.7)	81.2 (8.8)	82.0 (9.7)	83.3 (19.5)	83.4 (8.3)
HOMA-IR, median (IQR)	1.4 (1.1-1.9)	1.4 (1.1-1.8)	1.5 (1.2-2.0)	1.7 (1.3-2.2)	1.8 (1.4-2.3)
Total cholesterol, mg/dL	175.9 (32.5)	175.0 (32.4)	177.6 (32.0)	176.1 (32.5)	180.8 (38.4)
Triglycerides, mg/dL	67.4 (35.6)	66.0 (33.8)	70.3 (40.7)	67.8 (33.2)	73.5 (39.6)
Gamma-glutamyl transferase, U/L	22.8 (17.2)	21.1 (15.9)	25.2 (18.7)	26.2 (18.8)	30.8 (20.4)
FLI, median (IQR)	1.4 (1.1-1.8)	1.5 (1.2-2.0)	1.7 (1.3-2.2)	1.8 (1.4-2.3)	1.4 (1.1-1.8)
C-reactive protein at Y7, median (IQR), mg/L	1.0 (0.5-2.9)	0.9 (0.4-2.2)	1.4 (0.6-3.8)	1.7 (0.7-4.1)	1.9 (0.8-3.6)
N of cardiometabolic risk factors at Y25b	2.3 (1.5)	1.7 (1.2)	3.2 (1.3)	3.6 (1.1)	3.5 (1.3)
≥1, %	88.9	82.9	99.4	100.0	100.0
Antihypertension medication use over 25 yrs, %	26.5	0.3	62.7	89.2	94.7
25-year changec
Alcohol intake, drink/wk	0.3 (8.3)	0.2 (8.2)	0.4 (8.7)	0.05 (7.3)	0.6 (10.4)
Weight, kg	17.0 (13.7)	14.9 (12.7)	19.3 (13.9)	23.5 (15.1)	21.5 (14.8)
BMI, kg/m2	5.9 (4.8)	5.1 (4.4)	6.8 (4.9)	8.2 (5.4)	7.5 (5.2)
Fasting glucose, mg/dL	15.6 (19.9)	12.9 (14.4)	19.6 (26.1)	21.5 (25.3)	22.5 (32.0)
Total cholesterol, mg/dL	16.8 (35.9)	19.8 (32.8)	12.4 (39.8)	10.9 (38.4)	4.6 (50.5)

Data are mean (SD) otherwise indicated. Characteristics are shown at Y0 unless noted; family income was first assessed at Y5, and C-reactive protein level at Y7.

^aParticipants with missing values for family income (n = 203, 7.2%) was categorized into a separate “missing” category. All other variables are complete.

^bCardiometabolic risk factors included: 1) BMI ≥25 kg/m² or waist circumference >94 cm (for men) or 80 cm (for women); 2) fasting glucose ≥100 mg/dL or 2-hour post-load glucose ≥140 mg/dL or hemoglobin A1C ≥5.7% or type 2 diabetes or treatment for type 2 diabetes; 3) systolic blood pressure ≥130 mmHg or diastolic blood pressure ≥85 mmHg or specific antihypertensive medication treatment; 4) triglycerides ≥150 mg/dL or lipid lowering medication treatment; 5) high-density lipoprotein cholesterol ≤40 mg/dL (in men) or 50 mg/dL (in women) or lipid lowering medication treatment.

^cThe 25-year change was calculated as the level at Y25 minus the level at Y0.

Abbreviations: IQR, interquartile range; BMI, body mass index; HOMA-IR, Homeostasis model assessment of insulin resistance; FLI, fatty liver index.

The prevalence of participants identified as having SLD at Y25 varied from 12.8% in the low-stable diastolic BP trajectory to 37.5% in the elevated-increasing trajectory. In fully adjusted Model 3, long-term diastolic BP trajectories were positively associated with SLD (Table 2). Compared with low-stable trajectory, the adjusted RRs were 2.02 (95% CI: 1.41-2.88) for elevated-increasing, 1.57 (95% CI: 1.21-2.04) for elevated-stable, 1.77 (95% CI: 1.33-2.35) for moderate-increasing, and 1.47 (95% CI: 1.17-1.84) for moderate-stable. Similar positive associations were observed for systolic BP trajectories with SLD. Participants with moderate-stable, moderate-increasing, and elevated-stable systolic BP trajectories had 43%, 63%, and 36% greater SLD risk, respectively, compared to those with low-stable systolic BP throughout young adulthood. Early-onset hypertension had a higher SLD risk compared to late-onset (32.9% vs. 31.8%). Specifically, the RRs were 1.75 (95% CI: 1.17-2.60) for early-onset, 1.75 (95% CI: 1.34-2.29) for mid-adulthood onset, and 1.62 (95% CI: 1.33-1.97) for late-onset hypertension, compared to normotensive participants. Consistently, we observed that faster annual BP increases were linked to higher SLD risk, with RRs per 1 mmHg/year of 1.61 and 1.25 for diastolic BP and systolic BP, respectively (Table 3). Associations were even stronger among individuals with normal BP over 25 years, with RRs of 2.42 (95% CI: 1.85-3.16) for diastolic BP and 1.49 (95% CI: 1.17-1.89) for systolic BP per 1 mmHg/year increase.

Table 2.

Adjusted relative risks and 95% confidence intervals of the associations between blood pressure and hypertension trajectories for 25 years with steatotic liver disease

Trajectories	No. of SLD at Y25 (% of N)	RR (95% CI)
		Model 1	Model 2	Model 3
Diastolic BP
Low-Stable	82 (12.8)	1 (Ref)	1 (Ref)	1 (Ref)
Moderate-Stable	285 (22.0)	1.78 (1.41-2.24)	1.53 (1.22-1.92)	1.47 (1.17-1.84)
Moderate-Increasing	83 (28.9)	2.62 (1.98-3.46)	2.06 (1.56-2.72)	1.77 (1.33-2.35)
Elevated-Stable	139 (27.5)	2.31 (1.79-2.98)	1.85 (1.43-2.38)	1.57 (1.21-2.04)
Elevated-Increasing	33 (37.5)	3.44 (2.44-4.84)	2.63 (1.88-3.68)	2.02 (1.41-2.88)
Systolic BP
Low-Stable	94 (12.5)	1 (Ref)	1 (Ref)	1 (Ref)
Moderate-Stable	309 (23.7)	1.83 (1.47-2.28)	1.54 (1.24-1.92)	1.43 (1.15-1.77)
Moderate-Increasing	66 (27.8)	2.53 (1.91-3.36)	1.96 (1.48-2.60)	1.63 (1.22-2.18)
Elevated-Stable	133 (29.2)	2.12 (1.64-2.74)	1.67 (1.30-2.14)	1.36 (1.05-1.76)
Elevated-Increasing	20 (30.8)	2.70 (1.78-4.08)	2.03 (1.34-3.06)	1.49 (0.98-2.28)
Hypertension
Mostly normotensive	306 (16.9)	1 (Ref)	1 (Ref)	1 (Ref)
Late-onset hypertension	197 (31.8)	1.98 (1.70-2.31)	1.69 (1.44-1.97)	1.62 (1.33-1.97)
Mid-adulthood onset hypertension	94 (29.8)	2.23 (1.83-2.71)	1.87 (1.53-2.27)	1.75 (1.34-2.29)
Early-onset hypertension	25 (32.9)	2.22 (1.60-3.08)	1.87 (1.32-2.64)	1.75 (1.17-2.60)

Hypertension trajectories refer to a risk pattern rather than a definitive progression to hypertension for all individuals in this group.

Model 1: Modified Poisson regression adjusting for age (years, continuous), sex (women, men), self-reported race (Black, White), field center (Birmingham, Chicago, Minneapolis, Oakland);

Model 2: model 1 plus education (high school or less, some college, college graduate or higher), family income (less than $25,000, $25,000 through $50,000, $50,000 and greater, missing), smoking status (never, former, current), alcohol intake (never/former, current: ≤7(women)/14(men) drinks/week, current: >7(W)/14(M) drinks/week), physical activity (exercise units, in quartiles), the healthy eating index-2015 (in quartiles), body mass index (kg/m², continuous), fasting glucose (mg/dL, continuous), total cholesterol level (mg/dL, continuous) at Y0;

Model 3: model 2 plus use of antihypertension medications over Y0 to Y25 (yes, no).

Abbreviations: SLD, steatotic liver disease; RR, relative risk; 95% CI: 95% confidence interval; BP, blood pressure.

Table 3.

Adjusted relative risks and 95% confidence intervals of the associations between annual blood pressure increase rates with steatotic liver disease

Population	Annual increase rate
	Diastolic BP	Systolic BP
Total population (n=2817)
Mean (SD), mmHg/year	0.31 (0.58)	0.47 (0.46)
RR (95% CI)	1.61 (1.40-1.85)	1.25 (1.12-1.40)
Participants with normal BP over 25 years (n=1784)
Mean (SD), mmHg/year	0.21 (0.37)	0.32 (0.42)
RR (95% CI)	2.42 (1.85-3.16)	1.49 (1.17-1.89)

Modified Poisson regression adjusting for age (years, continuous), sex (women, men), self-reported race (Black, White), field center (Birmingham, Chicago, Minneapolis, Oakland), education (high school or less, some college, college graduate or higher), family income (less than $25000, $25000 through $50000, $50000 and greater, missing), smoking status (never, former, current), alcohol intake (never/former, current: ≤7(women)/14(men) drinks/week, current: >7(W)/14(M) drinks/week), physical activity (exercise units, in quartiles), the healthy eating index-2015 (in quartiles), body mass index (kg/m², continuous), fasting glucose (mg/dL, continuous), and total cholesterol level (mg/dL, continuous) at Y0.

Abbreviation: BP, blood pressure; SD standard deviation; RR, relative risk; 95% CI, 95% Confidence interval.

In the current study, 79.1% of SLD cases were classified as MASLD, while 20.9% were attributed to other SLD subtypes (7.7% MetALD, 3.5% ALD, and 9.6% due to other causes). BP trajectories were significantly linked to both subtypes (Table S7). Stronger associations were observed for moderate-to-severe SLD (Table S8), with the elevated-increasing diastolic BP trajectory showing approximately threefold risk (RR: 2.87, 95% CI: 1.54-5.36) compared with low-stable trajectory.

In sensitivity analyses, associations between BP and hypertension trajectories and risk of developing SLD were essentially unchanged when using 20-year trajectories from Y0 to Y20 (Table S9), after additional adjustment for waist circumference in place of BMI, HOMA-IR in place of fasting glucose, other lipid metrics (low-density lipoprotein, high-density lipoprotein cholesterol, triglycerides) in place of total cholesterol, CRP, or chronic kidney disease over 25 years (Table S10). Further adjustment for 25-year changes in alcohol intake, BMI, fasting glucose, and total cholesterol slightly attenuated associations. Results were consistent when restricted to participants without antihypertensive medication use, self-reported liver disease, with a FLI <30 at Y0, or after further excluding lean SLD cases (Table S11). Excluding participants with discordant BP trajectories slightly reduced effect sizes, while associations remained robust across alternative hypertension definitions (Table S12–S13). The E-value analysis³² showed that an unmeasured or unadjusted confounder would need to be associated with both BP or hypertension trajectories and SLD at RR ranging from 2.06 to 3.46 to mitigate the observed associations (Table S14), far stronger than measured powerful confounders such as BMI (RR: 1.11, 95% CI: 1.09-1.13) or antihypertensive medication use (RR: 1.44, 95% CI: 1,18-1.76). It reinforces the robustness of the observed association against residual confounding.

The overall trajectories are similar between men and women, although the elevated-increasing SBP trajectory appeared to rise more steeply among women (Figure S1). When further examined in relation to SLD, the associations were generally consistent across sex (Table S15). Stratified analyses indicate that patterns of associations were similar across self-reported race, healthy lifestyle score, BMI, and genetic susceptibility, with interaction tests generally not showing statistical evidence of difference (Tables S16–S18).

Discussion

In a 25-year U.S. cohort of Black and White adults, we identified five trajectories for diastolic BP and systolic BP, and four hypertension trajectories from young adulthood to midlife. Our findings demonstrate that individuals with rapidly increasing BP trajectories, as well as those with early onset of hypertension, exhibited significantly higher midlife SLD risk. Consistently faster BP annual increases were also associated with greater SLD risk, particularly among those with normal BP during follow-up. These findings add evidence that maintaining a low and stable BP patterns are linked to lower lifetime SLD risk.

To our knowledge, ours is the first longitudinal observational study examining associations between BP trajectories over decades with prevalent SLD. Our findings are consistent with previous studies linking higher BP from single-visit measures to SLD development and progression.^8–11 Our results further suggest that both BP level and rate of increase, even below the hypertension threshold, were associated with SLD risk, independent of treated hypertension. Certain populations are more likely to experience rapid BP increases, placing them at higher SLD risk. Diastolic BP trajectories were more strongly associated with SLD than systolic BP trajectories in our study, possibly due to reduced peripheral amplification of systolic BP from pulse wave reflection in young adults may obscure detection of moderately elevated central systolic BP in this age group.³³ Excluding participants with discordant diastolic and systolic BP trajectories did not materially change our finding. Additionally, earlier hypertension onset was associated with a higher SLD risk than late-onset. These findings support the notion that adverse effects of elevated BP and hypertension may take years to manifest, underscoring the importance of early-life risk assessment. Understanding the impact of timing and magnitude of BP changes on SLD risk could inform future interventional trials.

Our findings underscore the value of considering long-term BP trajectories in clinical and public health contexts. Trajectories capture life-course exposure variations and cumulative effects, minimizing misclassification from single-visit measurement. For instance, moderate-stable and moderate-increasing diastolic BP groups had similar low-risk initial BP level but different midlife SLD risk (RR 1.47 vs. 1.77), despite mean BP remaining below clinical hypertension thresholds over 25 years. This suggests that single-visit BP may overlook high-risk individuals, particularly those with rapid increases. Upcoming hypertension trials could incorporate BP trajectories as therapeutic and preventive targets to determine whether slowing BP increases in young adulthood benefits liver health.

In the present study, we modeled BP trajectories across entire population to maximize statistical power and ensure the robustness of trajectory modeling. Although sexual dimorphism and hormonal influences on hypertension and SLD development are established,^34,35 our exploratory sex-specific analyses showed largely similar BP trajectories and associations with SLD in men and women, even after adjusting for women’s menopausal status. However, limited sample size restricted our ability to identify stable sex- or menopausal status-specific trajectories, particularly in increasing trajectories. Overall, our study did not provide evidence that varying BP patterns contribute to sex differences in SLD risk, though these results should be interpreted cautiously. In race-stratified analyses, while Black adults were more likely to be in increasing trajectories compared to White adults, the associations with SLD were more distinct among White adults, despite P for interaction is not statistically significant. These differences may be partly explained by the underrepresentation of Black participants, as 37.8% of CARDIA cohort lacked SLD assessment at Y25 and these individuals were more likely to be Black (58.1% Black vs. 41.9% White, Table S1). Furthermore, racial differences in SLD pathophysiology may play a role, as obesity and dyslipidemia playing a more prominent role than hypertension in Black individuals.³⁶ Larger-scale studies are warranted to validate these findings.

Both observational studies and MR analyses suggest that hypertension is a risk factor of SLD, but the underlying mechanisms remain incompletely understood.³⁷ Increased BP and SLD shared several pathophysiological pathways, particularly IR, which plays a central role in both conditions.^38,39 IR can enhance salt absorption, induce lipid peroxidation, and activate the secretion of proinflammatory cytokines.^13,40,41 These conditions can reduce vascular elasticity and luminal width, leading to BP increases and liver endothelial dysfunction, which may drive progression of simple SLD to steatohepatitis and fibrosis.⁴² Hypoadiponectinemia, an IR-independent risk factor for hypertension,⁴³ while adiponectin could reduce IR and attenuate hepatic inflammation and fibrosis.⁴⁴ Systemic inflammation may also link hypertension and SLD by activating the renin-angiotensin-aldosterone system and worsening IR.^45,46 Long-term BP variabilities are influenced by aging, altered baroreflex sensitivity, autonomic dysfunction, sleep disorders, excess dietary sodium, weight fluctuations, and poor drug adherence.^47–49 Animal studies suggest that BP changes may induce direct end-organ damage and endothelial dysfunction independent of mean BP level, through endothelial lesion, renin-angiotensin-aldosterone system activation, and systemic inflammation.^50,51 These processes can increase intrahepatic vascular resistance⁵² and stimulate fibroblast activation,⁵³ contributing to liver injury and fibrosis. Further studies are still needed to investigate the underlying mechanisms.

Strengths and limitations

Key strengths of this study include its well-characterized longitudinal cohort of Black and White Americans with repeated phenotyping over 25 years, enabling robust trajectory modeling. However, several limitations require comment. First, the use of Hawksley random-zero sphygmomanometer for early examinations systematically underestimates BP by approximately 1.5 mmHg, potentially misclassifying trajectory groups underestimating observed associations. Second, SLD was assessed using non-contrast CT rather than gold standard methods of liver biopsy or magnetic resonance-based techniques, which are impractical for large population-based studies.⁵⁴ CT is one of the validated approaches for SLD assessment in larger cohorts due to its non-invasive nature, high diagnostic accuracy, and wide availability. A major limitation is that CT does not assess liver fibrosis, which restricted our ability to examine BP trajectories in the context of advanced liver disease. Nevertheless, using stricter cutoffs (liver attenuation ≤40 HU) to capture moderate-to-severe SLD yielded stronger associations. Third, suspected SLD cases at Y0 were identified using FLI (≥60), a validated and widely used non-invasive screening tool for SLD in epidemiological studies and settings prior referring to further imaging accessment.⁵⁵ However, it may be confounded by fibrosis and lacks diagnostic strategy for intermediate values (FLI 30 to <60), potential leading to underdiagnosis of SLD.⁵⁵ But, our results were similar when limited to participants with FLI <30 at Y0 (ruling out SLD).²¹ Fourth, SLD subtypes were classified using self-reported alcohol intake, which may lead to misclassification. The small number of lean SLD cases (n = 36) precluded a robust analysis of this clinically important phenotype. Further excluding these cases yielded consistent results, indicating the robustness of our findings. Fifth, residual or unmeasured confounding may exist; specifically, antihypertensive medications use and its type, dose, and adherence could influence BP patterns. But our findings remained consistent even among participants who never used antihypertensive medications. Sixth, liver attenuation at Y25 was missing for nearly 1/3 of surviving CARDIA participants. Compared with those who had liver attenuation measured, those with missing data were more likely to be men and Black, have lower educational attainment, and carry a higher burden of cardiometabolic risk factors (i.e., current smoking, high fasting glucose and HOMA-IR, Table S1) at baseline. This differential attrition may introduce healthy participant selection bias and limit the generalizability of our findings. Finally, despite CARDIA’s racial and geographic diversity, the generalizability of our findings to other ethnic populations may be limited.

Conclusions

In this longitudinal study spanning over two decades, we found that elevated BP or rapidly increasing (even below the hypertensive threshold) were associated with a significantly higher SLD risk. Monitoring BP changes during young adulthood may provide additional insights to identify populations at higher SLD risk. Future studies should investigate whether modifying these BP trajectories can improve metabolic and hepatic outcomes.

Perspectives

SLD affects over 30% of adults and is now the leading cause of liver-related morbidities, making prevention a key public health priority. Hypertension is a known risk factor for SLD, yet most evidence relies on single BP measurements in midlife or later, overlooking dynamic BP exposure over time. The results of this study suggest that individuals with elevated BP or rapidly increasing BP trajectories had substantially higher midlife SLD risk. Faster BP increases were also associated with elevated SLD risk, especially among individuals free of hypertension over 25 years. Incorporating life-course BP patterns into clinical practice may improve identification of individuals at higher SLD risk and inform earlier preventive strategies.

Pathophysiological Novelty and Relevance.

What Is New?

This study is the first to examine associations between long-term BP changes from young adulthood and midlife SLD risk. We found that participants who experienced steadily increasing BP (moderate-increasing and elevated-increasing) or had an early onset of hypertension exhibited the greatest elevated risk.

What Is Relevant?

Hypertension is a known risk factor of SLD. Understanding early-life BP dynamics, including both timing and magnitude of changes, may substantially improve SLD risk stratification and enable more precise preventive interventions.

Clinical/Pathophysiological Implications?

Incorporating life-course BP changes into clinical practice could assist more accurate identification of individuals at risk of SLD. Launching preventions to reducing elevated BP and slowing its increase rate in young adulthood may lower the lifetime SLD risk, even in individuals who do not meet clinical diagnostic criteria for hypertension.

Non-standard Abbreviations and Acronyms

SLD

steatotic liver disease

BP

blood pressure

MR

mendelian randomization

BMI

body mass index

IR

insulin resistance

CARDIA

Coronary Artery Risk Development in Young Adults

FLI

fatty liver index

LA

liver attenuation

CT

computed tomography

HU

Hounsfield units

MASLD

metabolic dysfunction-associated steatotic liver disease

MetALD

metabolic dysfunction and alcohol-associated liver disease

ALD

alcohol-associated liver disease

CRP

c-reactive protein

HOMA-IR

homeostasis model assessment of insulin resistance

RR

relative risk

CI

confidence interval

Data Availability

Because of the sensitive nature of the data collected for this study, requests to access the data set from qualified researchers trained in human subject confidentiality protocols may be sent to the CARDIA Coordinating Center at coc@uab.edu.