Prognostic Value of Spatial Trajectory of J‐Point Amplitude Across Left Precordial Leads in a Community‐Based Population Authors: Chenhui Tai, Zhe Wang Published: 2026 ABSTRACT ABSTRACTIntroductionEarly repolarization has historically been regarded as benign in many settings, although specific J‐point features have been linked to adverse outcomes. Whether spatial variation in J‐point amplitude across adjacent precordial leads carries prognostic information remains unclear.MethodsThis retrospective cohort study included participants aged 40–90 years without self‐reported myocardial infarction or heart failure from the Third National Health and Nutrition Examination Survey (NHANES‐III). Latent class trajectory analysis was used to identify spatial trajectories of J‐point amplitude across leads V4–V6. The primary outcomes were all‐cause mortality and cardiac death. Cox proportional hazards regression and competing risk models were used to evaluate associations between J‐point trajectory patterns and outcomes.ResultsAmong 5696 eligible participants, the mean age was 57.6 ± 12.9 years, 3092 were women, and 3986 were non‐Hispanic White. LCTA identified two spatial trajectories: Class 1 included 5295 participants and showed a low, mildly decreasing amplitude pattern, whereas Class 2 included 401 participants and showed higher V4 amplitude with a steeper decline toward V5–V6. During a median follow‐up of 303 months, all‐cause deaths occurred in 2935 and 257 participants, with 790 and 87 cardiac deaths, respectively. In fully adjusted models, Class 2 was associated with higher risks of all‐cause mortality (HR, 1.20; 95% CI, 1.03–1.39) and cardiac death (HR, 1.31; 95% CI, 1.01–1.71).ConclusionA steep V4‐to‐V6 decline in J‐point amplitude was associated with increased long‐term all‐cause and cardiac mortality. This spatial ECG phenotype may capture prognostic information beyond conventional early repolarization criteria and warrants external validation. Abbreviations difference in J‐point amplitude between leads action potential Bayesian Information Criterion body mass index classification and regression tree confidence interval cardiovascular disease diabetes mellitus electrocardiogram early repolarization pattern hazard ratio International Classification of Diseases, Tenth Revision transient outward current long‐acting injectable latent class trajectory analysis left ventricular hypertrophy National Center for Health Statistics National Death Index Third National Health and Nutrition Examination Survey negative predictive value positive predictive value poor R‐wave progression corrected QT interval standard deviation subdistribution hazard ratio Introduction The standard 12‐lead electrocardiogram (ECG) remains a widely available and low‐cost tool for cardiovascular assessment in both clinical and population settings (Friedman 2024). Beyond its diagnostic role in conduction disease, ischemia, and arrhythmias, ECG‐derived markers such as QRS duration, QT interval, voltage criteria for left ventricular hypertrophy (LVH), and ST‐T abnormalities have been associated with cardiovascular morbidity and mortality in community‐based cohorts (Lee et al. 2024; Tikkanen et al. 2022). However, conventional ECG‐based risk stratification has largely relied on global intervals, categorical thresholds, or visually defined abnormalities, while the spatial distribution of subtle waveform amplitudes across adjacent leads remains less explored (Malik 2024). The J point marks the junction between the end of ventricular depolarization and the beginning of the ST segment on the surface ECG. J‐point elevation and J‐wave manifestations have been linked to dispersion of repolarization within the ventricular myocardium (Hlaing et al. 2005). Because J‐point amplitude is routinely captured on standard ECGs, it can be quantified at the population level, including in individuals without visually apparent ECG abnormalities. Its amplitude varies across leads according to cardiac orientation, regional ventricular activation, early repolarization gradients, and the projection of electrical forces onto the body surface (Nademanee et al. 2026). Thus, beyond its traditional interpretation as part of a discrete ECG pattern, J‐point amplitude may be conceptualized as a continuous and spatially ordered electrophysiological signal. Historically, interest in the J point has centered on early repolarization pattern (ERP), typically defined by J‐point elevation above a fixed threshold in contiguous inferior or lateral leads, often accompanied by terminal QRS notching or slurring, with or without ST‐segment elevation (Nademanee et al. 2026). ERP is common in the general population and has often been considered benign, particularly among young adults, men, and athletes (Biasco et al. 2013; Ji et al. 2021; Klatsky et al. 2003). Nevertheless, certain J‐point features have been associated with idiopathic ventricular fibrillation, sudden cardiac death, and adverse cardiovascular outcomes, especially when present in inferior leads or accompanied by horizontal or descending ST segments (Aizawa et al. 2013; Haruta et al. 2011; Naruse et al. 2012). Other population‐based studies, however, have reported inconsistent associations (Ilkhanoff et al. 2014; Rollin et al. 2012). These discrepancies may reflect differences in population characteristics, lead territories, ST‐segment morphology, amplitude thresholds, and case definitions, highlighting the limitations of dichotomizing J‐point information using fixed cutoffs. Importantly, J‐point amplitude may carry prognostic information even below conventional thresholds for ERP. The clinical relevance of J‐point abnormalities also appears to depend partly on their lead distribution. For example, posterior J waves detected by synthesized V7–V9 leads and body surface mapping were common in patients with early repolarization syndrome and may be related to arrhythmogenesis (Miyamoto et al. 2023). Although these approaches are not routinely available in population cohorts, V4–V6 are standard, anatomically contiguous left precordial leads that provide a practical window into regional left ventricular electrical activity. Modeling J‐point amplitude across V4–V6 may therefore capture spatial heterogeneity that is missed by conventional threshold‐based definitions. Accordingly, we applied latent class trajectory analysis (LCTA) to identify spatial trajectories of J‐point amplitude across V4–V6. We then examined whether these trajectory‐defined ECG phenotypes were associated with long‐term all‐cause mortality and cardiac death in a nationally representative cohort. This approach evaluates interlead variation in J‐point amplitude as a population‐level ECG phenotype, rather than reclassifying individuals solely by conventional early repolarization criteria. Methods The Third National Health and Nutrition Examination Survey III (NHANES III) is a nationally representative survey conducted by the National Center for Health Statistics (NCHS) to assess disease prevalence, cardiovascular risk factors, and nutritional status among the civilian, noninstitutionalized U.S. population. The NCHS is part of the Centers for Disease Control and Prevention and conducts statistical and epidemiological activities under the authority of the Public Health Service Act. NHANES data are protected by federal confidentiality regulations. The NHANES III protocol was reviewed and approved by the NCHS Research Ethics Review Board, and all participants provided written informed consent. All data used in this study are publicly accessible at https://www.cdc.gov/nchs/nhanes/. Study Population In NHANES III, 6999 participants aged 40–90 years with normal sinus rhythm and no evidence of bundle branch block or intraventricular conduction delay were initially eligible. Participants were excluded if they had self‐reported heart failure, self‐reported myocardial infarction, cancer, or unreliable J‐point amplitude measurements. Sequentially, 315 participants with heart failure, 339 with myocardial infarction, 644 with cancer, and 5 with unreliable J‐point amplitude data were excluded. After these exclusions, 5696 participants were included in the final analytic cohort. Participants were subsequently classified according to the LCTA‐derived spatial trajectory of J‐point amplitude across leads V4–V6 (Figure 1). Flow diagram of participant selection. Participants were selected from NHANES III among individuals with normal sinus rhythm and no evidence of bundle‐branch block or intraventricular conduction delay. After exclusion of participants with heart failure, myocardial infarction, cancer, or missing J‐point amplitude data, 5696 participants were included in the final analysis and classified into two LCTA‐derived spatial trajectory classes of J‐point amplitude. Electrocardiography Standard 12‐lead ECGs were recorded by trained technicians during the NHANES III examination using the Marquette MAC 12 system (Marquette Medical Systems Inc.). ECG data were processed using computerized algorithms based on Minnesota and Nova coding systems (Blackburn et al. 1960; Rautaharju et al. 1998). The primary ECG exposure was defined using J‐point amplitudes measured in leads V4, V5, and V6. Additional ECG variables, including QT interval, QRS duration, poor R‐wave progression (PRWP), ECG‐defined LVH, and ERP, were considered as potential confounders or sensitivity‐analysis variables. The QT interval was corrected for heart rate using Bazett's formula: QTc = QT/√(RR) (Roguin 2011). QTc prolongation was defined as QTc ≥ 450 ms in men and ≥ 460 ms in women. PRWP was defined as an R‐wave amplitude ≤ 0.3 mV in lead V3 and the R‐wave amplitude in lead V2 being less than or equal to the R‐wave amplitude in lead V3. ECG‐defined LVH was defined using the Sokolow–Lyon voltage criterion (SV1 + RV5 or RV6 ≥ 3.5 mV) (Sokolow and Lyon 1949). ERP was defined as J‐point elevation of at least 0.1 mV in two or more contiguous inferior and/or lateral leads in the presence of a narrow QRS complex, defined as QRS duration < 120 ms. Leads V1–V3 were excluded from this definition to avoid overlap with Brugada‐type ECG patterns (Macfarlane et al. 2015). Because detailed terminal QRS morphology, including notching or slurring, was not uniformly available, this definition was based primarily on J‐point elevation and QRS duration, consistent with prior consensus recommendations (Antzelevitch and Di Diego 2022). Adjacent‐lead J‐point amplitude differences were calculated as ΔJ V5–V4 and ΔJ V6–V5 to quantify local spatial changes in J‐point amplitude across the V4‐to‐V6 axis. These indices were used to describe the trajectory pattern and were additionally included as ECG covariates in the fully adjusted model. Covariates Demographic characteristics, including age, sex, race/ethnicity, and household income, were obtained through household interviews. Race/ethnicity was categorized as non‐Hispanic White, non‐Hispanic Black, and other race/ethnicity. Household income was categorized as < $20,000/year or ≥ $20,000/year. Smoking status was assessed by questionnaire and classified as never, former, or current smoking. Body mass index (BMI) was calculated as weight in kilograms divided by height in meters squared. Self‐reported medical history included diabetes mellitus (DM) and stroke. Family history of CVD, use of antihypertensive medication, and use of medications that may influence J‐point characteristics were also recorded. Based on previous electrophysiological and clinical evidence, J‐point–influencing medications included antiarrhythmic agents, antianginal drugs, and psychotropic medications (Antzelevitch 2008). Blood pressure was calculated as the average of six measurements, including three obtained during the household interview and three obtained at the mobile examination center. Ascertainment of Mortality Outcomes The primary outcome of this study was all‐cause mortality. Mortality data were obtained by linking to the National Death Index, a publicly available dataset in the United States as of December 31, 2019. This database was used to determine the mortality status of eligible participants. According to the tenth revision of the International Classification of Diseases (ICD‐10), data on potential causes of death were used for case definition. Cardiac death was defined as death due to heart disease (ICD‐10 codes I00–I09, I11, I13, and I20–I51). Statistical Analyses Statistical analyses and data visualization were performed using R software, version 4.2.2. Continuous variables are presented as mean ± standard deviation or median with interquartile range, as appropriate. Categorical variables are presented as counts and percentages. Between‐group comparisons were performed using the t‐test or Mann–Whitney U test for continuous variables and the chi‐square test or Fisher's exact test for categorical variables, as appropriate. Missing covariate data were handled using multiple imputation. LCTA was used to identify distinct spatial trajectories of J‐point amplitude across the left precordial leads V4–V6. Although LCTA is commonly applied to longitudinal data, the same modeling framework can be applied to ordered non‐temporal dimensions, including spatial or anatomical sequences. In the present study, the ordered sequence of leads V4, V5, and V6 was treated as a spatial measurement axis to characterize interlead transitions in J‐point amplitude. This approach is consistent with prior methodological work on latent class models for ordered or spatially structured measurements (Proust‐Lima et al. 2017; Wall and Liu 2009). Initially, we fitted single‐class models with different functional forms for the spatial trajectory of J‐point amplitude across leads V4–V6, including linear and quadratic functions of lead order. The optimal functional form was selected based on the Bayesian Information Criterion (BIC). To avoid overfitting driven solely by BIC, clinically plausible trajectory patterns were also considered, and only latent classes comprising at least 5% of the study population were retained. Posterior probabilities of class membership were calculated for each participant, and individuals were assigned to the class with the highest posterior probability. To improve the clinical interpretability and reproducibility of the LCTA‐derived classes, we additionally developed a simplified classification and regression tree (CART) to approximate LCTA class membership using directly measurable ECG indices from leads V4–V6. The LCTA‐derived class was used as the reference outcome. Candidate predictors included J‐point amplitudes in leads V4, V5, and V6, the V6–V4 J‐point amplitude difference, the V4–V6 slope, and a curvature index. The final decision tree was fitted in the full analytic sample for visualization, whereas its classification performance was internally evaluated using 10‐fold cross‐validation. Sensitivity, specificity, positive predictive value, negative predictive value, accuracy, and balanced accuracy were calculated, with LCTA‐derived Class 2 treated as the positive class. Cox proportional hazards regression models were used to evaluate the association between LCTA‐derived J‐point amplitude class and all‐cause mortality. Hazard ratios and 95% confidence intervals were estimated using Class 1 as the reference group. Multivariable models were sequentially adjusted for demographic characteristics, clinical risk factors, medication use, and ECG covariates, as specified in the corresponding tables. For cardiac death, Fine–Gray subdistribution hazard models were used to account for competing risks from noncardiac death. These models were implemented using the cmprsk package in R. Subdistribution hazard ratios and 95% confidence intervals were reported. Subgroup analyses were conducted according to sex, age, race/ethnicity, body mass index, diabetes status, use of J‐point–influencing medications, ECG‐defined LVH and QT prolongation. Interaction p values were calculated to assess whether the association between LCTA‐derived class and mortality differed across subgroups. Sensitivity analyses were performed after sequential exclusion of participants with ECG‐defined LVH, use of J‐point–influencing medications, ERP, and QTc prolongation. A two‐sided p value < 0.05 was considered statistically significant. Results Study Population and Latent J‐Point Amplitude Classes Baseline characteristics according to LCTA‐derived J‐point amplitude trajectory class are shown in Table 1. The mean age of the study population was 57.58 ± 12.92 years; 3092 participants were women (54.3%), and 3986 were White (70.0%). Class 1 included 5295 participants (93.0%), whereas Class 2 included 401 participants (7.0%). During a median follow‐up of 303 months, 3192 all‐cause deaths occurred (56.0% of the cohort), including 877 cardiac deaths (15.4% of the cohort; 27.5% of deaths). TABLE 1Baseline characteristics of study participants according to LCTA‐derived spatial trajectory class of left precordial J‐point amplitude.VariableAllClass 1Class 2 p N = 5696 N = 5295 N = 401Age, years57.58 ± 12.9257.48 ± 12.8258.90 ± 14.210.034Female, n (%)3092 (54.3)2994 (56.5)98 (24.4)< 0.001Race/ethnicity, n (%)White3986 (70.0)3758 (71.0)228 (56.9)< 0.001Black1524 (26.8)1369 (25.9)155 (38.7)Other race/ethnicity186 (3.3)168 (3.2)18 (4.5)Total annual income < $20,000, n (%)2598 (46.4)2391 (45.9)207 (52.9)0.009Smoking status, n (%)Never2624 (46.1)2487 (47.0)137 (34.2)< 0.001Former1737 (30.5)1607 (30.3)130 (32.4)Current1335 (23.4)1201 (22.7)134 (33.4)Body mass index, kg/m2 27.69 ± 5.5227.81 ± 5.5426.10 ± 4.96< 0.001Stroke, n (%)143 (2.5)129 (2.4)14 (3.5)0.256Diabetes mellitus, n (%)592 (10.4)551 (10.4)41 (10.2)0.985Family history of CVD, n (%)671 (12.0)623 (12.0)48 (12.0)0.919Antihypertensive medications use, n (%)1509 (26.6)1388 (26.3)121 (30.2)0.100Diastolic blood pressure, mmHg78.02 ± 11.9477.90 ± 11.8179.71 ± 13.330.004Systolic blood pressure, mmHg132.89 ± 20.49132.43 ± 20.12138.88 ± 24.10< 0.001Use of J‐point–influencing drugs, n (%)464 (8.2)423 (8.0)41 (10.3)0.133QTc interval, ms430.19 ± 23.55430.41 ± 23.26427.27 ± 27.090.011QRS duration, ms95.32 ± 9.8295.16 ± 9.7797.45 ± 10.25< 0.001Poor R wave progression, n (%)217 (3.8)202 (3.8)15 (3.7)0.999ECG‐defined LVH, n (%)226 (4.0)173 (3.3)53 (13.3)< 0.001Early repolarization pattern, n (%)36 (0.6)5 (0.1)31 (7.7)< 0.001V1 J‐point amplitude, μV21.00 [5.00, 40.00]20.00[4.00, 37.00]45.00[20.00, 75.00]< 0.001V2 J‐point amplitude, μV47.00 [21.00, 80.00]45.00[20.00, 74.00]107.00[54.00, 151.00]< 0.001V3 J‐point amplitude, μV32.00 [6.00, 62.00]30.00[6.00, 57.00]99.00[39.00, 147.00]< 0.001V4 J‐point amplitude, μV16.00 [−6.00, 41.00]15.00[−6.00, 38.00]70.00[4.00, 116.00]< 0.001V5 J‐point amplitude, μV9.00 [−9.00, 28.00]9.00[−8.00, 27.00]14.00[−42.00, 62.00]0.344V6 J‐point amplitude, μV8.00 [−6.00, 22.00]8.00[−5.00, 22.00]5.00[−40.00, 40.00]0.004ΔJ amplitude, V5–V4, μV−6.00 [−18.00, 3.00]−5.00 [−16.00, 3.00]−50.00 [−64.00, −29.00]< 0.001ΔJ amplitude, V6–V5, μV−1.00 [−10.00, 7.00]−1.00 [−10.00, 7.00]−11.00 [−28.00, 7.00]< 0.001 Note: Values are expressed as mean ± SD, N (%), or median values [first quartile, third quartile]. Class 1 and Class 2 were identified using LCTA based on J‐point amplitude across leads V4–V6 and represent distinct spatial J‐point amplitude trajectories rather than predefined clinical categories.Abbreviations: ΔJ, difference in J‐point amplitude between adjacent leads; CVD, cardiovascular disease; ECG, electrocardiogram; LCTA, Latent class trajectory analysis; LVH, left ventricular hypertrophy; QTc, corrected QT interval. Baseline characteristics of study participants according to LCTA‐derived spatial trajectory class of left precordial J‐point amplitude. Note: Values are expressed as mean ± SD, N (%), or median values [first quartile, third quartile]. Class 1 and Class 2 were identified using LCTA based on J‐point amplitude across leads V4–V6 and represent distinct spatial J‐point amplitude trajectories rather than predefined clinical categories. Abbreviations: ΔJ, difference in J‐point amplitude between adjacent leads; CVD, cardiovascular disease; ECG, electrocardiogram; LCTA, Latent class trajectory analysis; LVH, left ventricular hypertrophy; QTc, corrected QT interval. A quadratic function provided the best fit for modeling the spatial trajectory of J‐point amplitude across leads V4–V6. LCTA identified two distinct J‐point amplitude trajectories. Class 1 included 5295 participants and was characterized by a relatively low and mildly decreasing J‐point amplitude trajectory across V4–V6. Class 2 included 401 participants and was characterized by a higher J‐point amplitude in V4 followed by a steeper decline toward V5 and V6. The class‐specific mean predicted trajectories are shown in Figure 2. A schematic representation of these two left precordial J‐point amplitude trajectories is provided in Figure 3 for illustrative purposes only and does not represent actual electrocardiographic tracings or ST‐segment morphology. The mean posterior probabilities of class assignment were 95.5% for Class 1 and 84.2% for Class 2, indicating acceptable classification certainty. Class‐specific mean predicted J‐point amplitude trajectories across leads V4–V6. Solid lines represent the latent class trajectory analysis–derived mean predicted trajectories of J‐point amplitude for each class across leads V4–V6. Dashed lines represent the corresponding 95% confidence intervals. A quadratic trajectory function was used to model the spatial distribution of J‐point amplitude across the left precordial leads. Schematic illustration of LCTA‐derived spatial trajectories of J‐point amplitude across the left precordial leads. The schematic illustrates the relative spatial pattern of J‐point amplitude across leads V4–V6 for Class 1 and Class 2. Class 1 represents relatively low J‐point amplitudes with a mild V4‐to‐V6 decline, whereas Class 2 represents higher V4 J‐point amplitude followed by a steeper decline toward V5 and V6. This figure is intended for conceptual illustration only and does not represent actual electrocardiographic recordings or ST‐segment morphology. Compared with participants in Class 1, those in Class 2 were slightly older, with a mean age of 58.90 ± 14.21 years versus 57.48 ± 12.82 years, respectively (p = 0.034), and had a lower body mass index, 26.10 ± 4.96 kg/m2 versus 27.81 ± 5.54 kg/m2 (p < 0.001). Class 2 included a substantially lower proportion of women than Class 1, with 98 women (24.4%) versus 2994 women (56.5%), respectively (p < 0.001). Race/ethnicity differed significantly between classes (p < 0.001); Class 2 had a lower proportion of White participants than Class 1, 228 (56.9%) versus 3758 (71.0%), and a higher proportion of Black participants, 155 (38.7%) versus 1369 (25.9%). Participants in Class 2 were more likely to have a total annual income < $20,000, 207 (52.9%) versus 2391 (45.9%) (p = 0.009), and were more likely to be current smokers, 134 (33.4%) versus 1201 (22.7%) (overall p for smoking status < 0.001). Systolic and diastolic blood pressure were also higher in Class 2 than in Class 1. The mean systolic blood pressure was 138.88 ± 24.10 mmHg versus 132.43 ± 20.12 mmHg, respectively (p < 0.001), and the mean diastolic blood pressure was 79.71 ± 13.33 mmHg versus 77.90 ± 11.81 mmHg, respectively (p = 0.004). Several ECG characteristics also differed significantly between the two classes. Participants in Class 2 had a longer QRS duration than those in Class 1, 97.45 ± 10.25 ms versus 95.16 ± 9.77 ms, respectively (p < 0.001), whereas the QTc interval was slightly shorter in Class 2, 427.27 ± 27.09 ms versus 430.41 ± 23.26 ms (p = 0.011). ECG‐defined LVH was more frequent in Class 2 than in Class 1, 53 participants (13.3%) versus 173 participants (3.3%) (p < 0.001). Early repolarization pattern was also more common in Class 2, occurring in 31 participants (7.7%) compared with 5 participants (0.1%) in Class 1 (p < 0.001). Importantly, the use of potentially J‐point–influencing drugs did not differ significantly between Class 2 and Class 1, 41 participants (10.3%) versus 423 participants (8.0%), respectively (p = 0.133). No significant between‐class differences were observed in the prevalence of stroke, diabetes mellitus, family history of CVD, antihypertensive medication use, or poor R‐wave progression; these variables are shown in Table 1. During follow‐up, all‐cause mortality occurred in 2935 participants in Class 1 (55.4%) and 257 participants in Class 2 (64.1%), corresponding to 790 (14.9%) and 87 (21.7%) cardiac deaths, respectively. Agent‐specific use of potentially J‐point–influencing medications is shown in Table S1. Overall, the most commonly used agents were verapamil, nifedipine, diltiazem, propranolol, and amitriptyline. Among participants in Class 2, the most frequent agents were nifedipine, verapamil, diltiazem, isosorbide dinitrate, and propranolol. Most individual agents were used infrequently in both classes. The prevalence of agents with antiarrhythmic properties, including verapamil, propranolol, procainamide, disopyramide, and propafenone, did not differ significantly between Class 1 and Class 2. Nifedipine use was more common in Class 2 than in Class 1, whereas other agent‐level comparisons were not statistically significant. Sodium‐channel–blocking antiarrhythmic agents were uncommon in the analytic cohort. Association of J‐Point Amplitude Trajectory Class With Mortality As shown in Table 2, Class 2 was associated with a significantly higher risk of all‐cause mortality in the unadjusted model (HR, 1.37; 95% CI, 1.20–1.55). The association remained significant after sequential adjustment for demographic and lifestyle factors in Model 2 (HR, 1.22; 95% CI, 1.07–1.39) and additional clinical covariates in Model 3 (HR, 1.15; 95% CI, 1.01–1.31). In the fully adjusted Model 4, which further accounted for ECG‐related covariates, including poor R‐wave progression, ECG‐defined LVH, ERP, QTc interval, QRS interval, and adjacent‐lead ΔJ indices, including ΔJ V5–V4 and ΔJ V6–V5, Class 2 remained independently associated with all‐cause mortality (HR, 1.20; 95% CI, 1.03–1.39). A similar pattern was observed for cardiac death. Class 2 was associated with a higher risk of cardiac death in the unadjusted model (HR, 1.56; 95% CI, 1.25–1.96). The association was attenuated but remained statistically significant after full adjustment (HR, 1.31; 95% CI, 1.01–1.71). TABLE 2Association between LCTA‐derived spatial trajectory class of J‐point amplitude and all‐cause and cardiac mortality.OutcomeLatent classModel 1Model 2Model 3Model 4HR (95% CI) p HR (95% CI) p HR (95% CI) p HR (95% CI) p All‐cause deathClass 11.00 (reference)—1.00 (reference)—1.00 (reference)—1.00 (reference)Class 21.37 (1.20–1.55)< 0.0011.22 (1.07–1.39)0.0031.15 (1.01–1.31)0.0371.20 (1.03–1.39)0.020Cardiac deathClass 11.00 (reference)—1.00 (reference)—1.00 (reference)—1.00 (reference)Class 21.56 (1.25–1.96)< 0.0011.40 (1.10–1.79)0.0061.29 (1.01–1.65)0.0441.31 (1.01–1.71)0.045 Note: Model 1 was unadjusted. Model 2 was adjusted for age, sex, race/ethnicity, income, body mass index, and smoking status. Model 3 was additionally adjusted for stroke, diabetes mellitus, family history of cardiovascular disease, antihypertensive medication use, systolic blood pressure, diastolic blood pressure, and use of J‐point–influencing drugs. Model 4 was further adjusted for poor R‐wave progression, ECG‐defined left ventricular hypertrophy, QTc interval, QRS duration, early repolarization pattern, and ΔJ amplitude (V6–V5 and V5–V4). Cox proportional hazards models were used for all‐cause mortality, and Fine–Gray competing‐risk models were used for cardiac death.Abbreviations: CI, confidence interval; ECG, electrocardiogram; HR, hazard ratio; LCTA, latent class trajectory analysis; QTc, corrected QT interval. Association between LCTA‐derived spatial trajectory class of J‐point amplitude and all‐cause and cardiac mortality. Note: Model 1 was unadjusted. Model 2 was adjusted for age, sex, race/ethnicity, income, body mass index, and smoking status. Model 3 was additionally adjusted for stroke, diabetes mellitus, family history of cardiovascular disease, antihypertensive medication use, systolic blood pressure, diastolic blood pressure, and use of J‐point–influencing drugs. Model 4 was further adjusted for poor R‐wave progression, ECG‐defined left ventricular hypertrophy, QTc interval, QRS duration, early repolarization pattern, and ΔJ amplitude (V6–V5 and V5–V4). Cox proportional hazards models were used for all‐cause mortality, and Fine–Gray competing‐risk models were used for cardiac death. Abbreviations: CI, confidence interval; ECG, electrocardiogram; HR, hazard ratio; LCTA, latent class trajectory analysis; QTc, corrected QT interval. Subgroup analyses for all‐cause mortality are shown in Figure 4. The association between Class 2 and all‐cause mortality was generally consistent across subgroups defined by age, sex, race/ethnicity, body mass index, diabetes status, ECG‐defined LVH, and QTc prolongation, with no statistically significant interactions for these variables. A significant interaction was observed for use of potentially J‐point–influencing drugs. The association between Class 2 and all‐cause mortality appeared stronger among medication users than among non‐users. In exploratory analyses according to medication category, the association was more pronounced among users of agents with antiarrhythmic properties than among non‐users, whereas no significant interaction was observed for antianginal drug use (Table S2). The interaction for psychotropic drug use did not reach statistical significance, and the corresponding estimate among psychotropic drug users was highly imprecise because of sparse data. The interaction for other drug categories was not estimable because of limited sample size. Sensitivity analyses for cardiac death are presented in Table 3. After excluding participants with ECG‐defined LVH, Class 2 remained associated with a higher risk of cardiac death. The association was also preserved after further exclusion of participants using potentially J‐point–influencing drugs and after additional exclusion of those with ERP. After further exclusion of participants with QTc prolongation, the point estimate remained directionally consistent but was no longer statistically significant. Subgroup analyses of the association between LCTA‐derived spatial trajectory class of J‐point amplitude and all‐cause mortality. Hazard ratios represent the risk of all‐cause mortality for Class 2 compared with Class 1 within each subgroup. Models were adjusted using the same covariates as the fully adjusted model unless otherwise specified. p values for interaction were obtained by including a cross‐product term between latent class and subgroup variable in the Cox proportional hazards model. BMI, body mass index; DM, diabetes mellitus; ECG, electrocardiogram; LVH, left ventricular hypertrophy; QTc, corrected QT interval. TABLE 3Sensitivity analyses for the association between LCTA‐derived spatial trajectory class of J‐point amplitude and cardiac death after sequential exclusion of potential confounders.Restricted population N HR (95% CI) p Excluding ECG‐defined LVH54701.38 (1.05–1.82)0.020Excluding ECG‐defined LVH and use of J‐point–influencing drugs50311.45 (1.09–1.93)0.011Excluding ECG‐defined LVH, use of J‐point–influencing drugs, and ERP50021.44 (1.08–1.92)0.013Excluding ECG‐defined LVH, use of J‐point–influencing drugs, ERP, and QTc prolongation43601.33 (0.94–1.87)0.110 Note: Estimates represent the association between Class 2 and cardiac death compared with Class 1 after sequential exclusion of participants with ECG‐defined LVH, use of potentially J‐point–influencing drugs, ERP, and QTc prolongation. Models were adjusted for the covariates included in the fully adjusted model, except for variables used to define the restricted population when applicable. Cardiac death was analyzed using Fine–Gray competing‐risk regression, with non‐cardiac death treated as a competing event. HR indicates subdistribution hazard ratio.Abbreviations: CI, confidence interval; ECG, electrocardiogram; ERP, early repolarization pattern; HR, hazard ratio; LCTA, latent class trajectory analysis; LVH, left ventricular hypertrophy; QTc, corrected QT interval. Sensitivity analyses for the association between LCTA‐derived spatial trajectory class of J‐point amplitude and cardiac death after sequential exclusion of potential confounders. Note: Estimates represent the association between Class 2 and cardiac death compared with Class 1 after sequential exclusion of participants with ECG‐defined LVH, use of potentially J‐point–influencing drugs, ERP, and QTc prolongation. Models were adjusted for the covariates included in the fully adjusted model, except for variables used to define the restricted population when applicable. Cardiac death was analyzed using Fine–Gray competing‐risk regression, with non‐cardiac death treated as a competing event. HR indicates subdistribution hazard ratio. Abbreviations: CI, confidence interval; ECG, electrocardiogram; ERP, early repolarization pattern; HR, hazard ratio; LCTA, latent class trajectory analysis; LVH, left ventricular hypertrophy; QTc, corrected QT interval. Simplified Decision Tree for Approximating LCTA‐Derived Class Membership To improve clinical interpretability, we developed a simplified CART using directly measurable ECG indices from leads V4–V6 to approximate LCTA–derived class membership. Class 2 was treated as the positive class. In 10‐fold cross‐validation, the decision tree identified Class 2 with a sensitivity of 94.8%, specificity of 91.0%, positive predictive value of 44.3%, negative predictive value of 99.6%, overall accuracy of 91.2%, and balanced accuracy of 92.9% (Figure S1 and Table S3). These findings suggest that the Class 2 spatial J‐point amplitude trajectory can be closely approximated using simple ECG‐based features, although external validation is required before clinical application. Discussion Our study demonstrates that the spatial trajectory of J‐point amplitude across the left precordial leads is associated with long‐term prognosis in adults with sinus rhythm. Using LCTA, we identified two distinct trajectories across V4–V6. Class 1 showed a relatively low and mildly decreasing amplitude trajectory, whereas Class 2 was characterized by a higher amplitude in V4 followed by a steeper decline toward V5 and V6. Compared with Class 1, Class 2 was associated with higher risks of both all‐cause mortality and cardiac death. These associations persisted after adjustment for demographic factors, cardiovascular risk factors, medication use, and ECG covariates, including poor R‐wave progression, ECG‐defined LVH, conventional ERP, QTc interval, QRS duration, and J‐point amplitude difference indices. These findings suggest that interlead variation in J‐point amplitude may provide prognostic information beyond conventional ECG thresholds and visually defined abnormalities. Previous studies have shown that the prognostic significance of early repolarization and related J‐point abnormalities depends on lead localization, J‐point amplitude, and accompanying ST‐segment morphology (Haïssaguerre et al. 2008). In population‐based cohorts, Tikkanen et al. (2009) and Sinner et al. (2010) further demonstrated that inferior lead localization and greater J‐point elevation were associated with increased mortality risk. Other studies have emphasized that horizontal or descending ST‐segment morphology after the J point identifies a higher‐risk phenotype (Tikkanen et al. 2011). Collectively, these findings indicate that the prognostic relevance of J‐point abnormalities is not determined solely by their presence, but by how their amplitude varies across leads and by the surrounding waveform context. The present study extends this literature in several ways. First, we did not restrict the analysis to participants meeting conventional early repolarization criteria. Instead, J‐point amplitude was evaluated as a continuous and spatially ordered ECG signal across anatomically contiguous left precordial leads. This approach may capture subthreshold variation that is overlooked by fixed cutoffs such as 0.1 or 0.2 mV. In our cohort, the J‐point amplitudes in both latent classes were generally below conventional early repolarization thresholds, suggesting that absolute amplitude alone may not fully capture risk. Second, the Class 2 phenotype should not be interpreted as equivalent to a horizontal or descending ST‐segment pattern. The latent classes were derived from interlead differences in J‐point amplitude rather than from ST‐segment slope within a single lead or terminal QRS notching or slurring. Thus, Class 2 represents a distinct spatial trajectory characterized by higher V4 J‐point amplitude and a steep decline toward V5–V6, rather than a conventional ST‐segment morphology. Third, although ERP and ECG‐defined LVH were more frequent in Class 2, the association between Class 2 and mortality persisted after accounting for these ECG features, supporting the incremental prognostic relevance of the spatial trajectory of J‐point amplitude. The mechanisms underlying the association between the Class 2 spatial trajectory of J‐point amplitude and adverse outcomes remain uncertain. Experimental evidence suggests that the electrocardiographic J wave reflects a transmural voltage gradient at the notch of the action potential (AP), resulting from a prominent transient outward current (I to) in the ventricular epicardium (Yan and Antzelevitch 1996; Litovsky and Antzelevitch 1988). Clinical electrophysiological data further indicate that regions exhibiting J waves may have shortened and spatially heterogeneous ventricular repolarization (Zhang et al. 2017). In this context, the Class 2 trajectory may reflect regional heterogeneity in terminal ventricular activation, early repolarization, or the projection of local electrical forces onto the body surface. Such heterogeneity could contribute to an arrhythmogenic substrate, although this interpretation remains mechanistic. At the same time, the observed spatial trajectory may also reflect myocardial remodeling. Participants in Class 2 had higher blood pressure and a higher prevalence of ECG‐defined LVH, suggesting that subclinical structural changes may have contributed to this phenotype (Saito et al. 2023). However, the association with mortality persisted after adjustment for ECG‐defined LVH and in sensitivity analyses excluding participants with ECG‐defined LVH, although unmeasured structural heart disease cannot be excluded in the absence of systematic cardiac imaging. Medication use may further modify, or alternatively mark, the risk associated with J‐point amplitude trajectory‐defined phenotypes. In subgroup analyses, the association between Class 2 and all‐cause mortality appeared stronger among users of potentially J‐point–influencing medications, particularly agents with antiarrhythmic properties. This finding is biologically plausible because drugs affecting depolarization, repolarization, ion‐channel activity, or autonomic tone may influence J‐point expression and repolarization heterogeneity (Alameh et al. 2026). Prior clinical observations have also suggested that sodium‐channel–blocking agents may accentuate arrhythmia risk in susceptible patients with J‐point or early repolarization phenotypes (Lin et al. 2015). However, these findings should be interpreted cautiously because most individual agents were infrequently used, sodium‐channel–blocking antiarrhythmic drugs were uncommon, and medication use may reflect confounding by indication. Moreover, pre‐treatment ECGs were unavailable; therefore, we cannot determine whether the Class 2 pattern preceded medication exposure or was modified by therapy. From a clinical perspective, this study suggests that the spatial trajectory of J‐point amplitude across V4–V6 may represent a feasible ECG phenotype for population‐level risk assessment because these leads are routinely available on standard 12‐lead ECGs. The decision‐tree model developed in this study further suggests that latent trajectory class membership can be approximated using simple, directly measurable ECG features, which may improve interpretability and facilitate future validation. However, this model should not yet be considered a clinical decision tool, and its incremental value over existing ECG and clinical risk markers requires confirmation in independent cohorts. This study has several limitations. First, its observational design precludes causal inference. Second, although participants with known myocardial infarction, heart failure, cancer, bundle branch block, or intraventricular conduction delay were excluded, echocardiography, cardiac magnetic resonance imaging, and histological data were unavailable; therefore, occult structural heart disease, subclinical hypertrophy, or cardiomyopathy could not be fully excluded. Third, detailed terminal QRS morphology and ST‐segment slope were not uniformly available, limiting direct comparison with traditional early repolarization phenotypes. Fourth, medication data lacked information on dose, duration, indication, and ECG findings before treatment initiation. Fifth, J‐point amplitude was assessed from a single baseline ECG, and changes over time in the V4–V6 spatial trajectory could not be evaluated. Finally, both the latent trajectory phenotype and the simplified decision‐tree model require external validation before clinical implementation. Conclusion In adults with sinus rhythm, a steep V4‐to‐V6 decline in J‐point amplitude was associated with increased risks of long‐term all‐cause mortality and cardiac death. This phenotype was not captured by conventional early repolarization criteria or fixed J‐point elevation thresholds. A simplified decision‐tree model may facilitate its identification using routine ECG measures, although external validation is required. Author Contributions Chenhui Tai: Conceptualization, methodology, data collection, data analysis and interpretation, and drafting and revising the manuscript. Zhe Wang: Conceptualization, methodology, supervision, data interpretation, critical revision of the manuscript, and final approval of the manuscript. All authors have approved the final version and agree to be accountable for all aspects of the work in accordance with the ICMJE authorship criteria. Funding The authors have nothing to report. Ethics Statement The NHANES study protocols were approved by the Research Ethics Board of the National Center for Health Statistics at the Centers for Disease Control and Prevention (CDC). Additionally, the research protocol received approval from the Institutional Review Board of the Nantong First People's Hospital. Consent Written informed consent was obtained from all participants in the NHANES III study. This secondary analysis used publicly available, de‐identified data. All authors approved the final version of the manuscript and consented to its publication. Conflicts of Interest The authors declare no conflicts of interest. Supporting information Table S1: Agent‐specific use of potentially J‐point–influencing medications according to LCTA‐derived spatial trajectory class of J‐point amplitude. Table S2: Exploratory subgroup analyses of the association between latent J‐point amplitude trajectory class and all‐cause mortality according to categories of potentially J‐point–influencing medication use. Table S3: Internal cross‐validated performance of a simplified decision tree for identifying latent trajectory Class 2. Figure S1: Simplified decision tree for approximating LCTA‐derived J‐point trajectory class.