跳到正文
原文
Nature 健康护理主题·· 18 天前AI 评分37

全口牙缺失与成人心理压力及死亡率关系研究

Frequent psychological distress and mortality among adults with all tooth loss across health-risk profiles

AI 导读

一项基于美国国家健康访谈调查数据的研究发现,全口牙缺失(ATL)与成人频繁负面情绪(FNE)相关,且FNE与全因及特定原因死亡率增加有关。研究通过潜在类别分析识别健康风险特征(HRPs),并评估HRPs与FNE的关系。结果表明,HRPs与FNE相关,而FNE在HRPs基础上进一步增加了全因死亡率的风险,提示心理压力可能是全口牙缺失成人死亡率的额外影响因素。

正文

Introduction

Oral diseases, such as periodontitis, are highly prevalent, particularly among older adults1. Chronic inflammation associated with oral diseases has been linked to tooth loss, which is associated with.

multiple adverse health outcomes, such as atherosclerosis and cardiovascular events2. Tooth loss may also influence diet, nutrient intake, and food choices, potentially contributing to malnutrition and mortality3. Consequently, all tooth loss (ATL) may reflect poorer overall health status. In addition, ATL is commonly associated with aging, chronic diseases, and health-related behaviors, including diabetes, hypertension, and smoking, which may also be related to mortality among adults with ATL4. Therefore, a better understanding of mortality patterns and associated factors among adults with ATL is warranted.

Healthy-risk profiles (HRPs), including hypertension, diabetes, and smoking, are associated with long-term outcomes among adults with ATL. For example, diabetes has been linked to complications affecting multiple organs and elevated mortality. However, adults with similar HRPs often show heterogeneous outcomes, suggesting that additional factors may influence health outcomes among adults with ATL. ATL can affect facial appearance and pronunciation, and may limit social activities, potentially contributing to psychological distress and depression, which may be related to overall physiological and psychological health5. Conversely, ATL tends to occur in individuals with poorer health status, which may also be associated with negative emotions6. Previous studies have reported bidirectional associations between oral health and depression, with depression potentially affecting oral and overall health7. Therefore, negative emotions may represent an additional factor associated with mortality among adults with ATL, beyond HRPs.

We explored the association between ATL and frequent negative emotion (FNE) using adult.

data from the National Health Interview Survey of US (NHIS). Latent class analysis (LCA) was performed to identify descriptive, LCA-derived HRPs based on baseline health-risk indicators. Compared with individual disease variables, these HRPs were intended to summarize co-occurring health-risk patterns, rather than to define reproducible biological or clinical subtypes. The associations of HRPs with FNE were also examined. We further assessed the associations of FNE with all-cause and cause-specific mortality across HRP groups. Time-dependent receiver operating characteristic (ROC) curve analysis was conducted as an exploratory assessment of the incremental discriminatory information provided by adding FNE to a model that included HRPs for all-cause mortality. Exploratory, association-based population attributable fractions (PAFs) associated with FNE for all-cause mortality among adults with ATL were also estimated.

Materials and methods

Study design and population

A conceptual framework was developed to guide the study design and statistical analyses. Negative emotion was hypothesized to be associated with mortality through multiple potential pathways. HRPs were considered to reflect the cumulative burden of multimorbidity, lifestyle, and metabolic status, which may influence the studied associations. Covariates were selected based on their relationships with negative emotion, health status, and mortality. As a population-based observational study, the results should be interpreted as associations rather than causal effects.

We used adult data of NHIS from 1998 to 2018, which enrolled a representative sample of US households. The NHIS is performed by the National Center for Health Statistics of the Centers for Disease Control and Prevention. NHIS collected self-reported data on health topics. To obtain population-based estimates, each adult was assigned a survey weight, which was composed of a base sampling weight and adjustments for both screener and topical nonresponse, selection of a single adult, and characteristics of adult across different states. In the present study, survey weights were applied in descriptive and univariable analyses. Adults were defined as individuals aged 18 years or older. To improve model stability and reduce instability due to sparse data at advanced ages, age was truncated at 85 years for participants aged 85 years and older, a common approach in population-based epidemiologic analyses8.

Inclusion criteria: adults with complete ATL data (with or without ATL) from 1998 to 2018 in the NHIS database were included. ATL was defined as loss of both upper and lower teeth. The recruited sociodemographic covariates included interview year, age, age group, sex, house region, race, marital status, body mass index (BMI), BMI category, smoking status, and alcohol consumption status. The disease history variables included self-reported history of hypertension, coronary heart diseases (CHD), stroke, emphysema, asthma, cancer, and diabetes.

Exclusion criteria: lack of follow-up data, lack of detailed cause of death, lack of ATL data, lack of complete sociodemographic covariates, or disease history information.

The study was exempt from ethic review by the Institutional Review Board because of the use of publicly available deidentified data (https://www.cdc.gov/nchs/nhis/documentation/index.html). This study also followed the Strengthening the Reporting of observational studies in Epidemiology (STROBE) reporting guidelines9.

Baseline health-risk indicators used to derive HRPs

We selected 10 baseline health-risk indicators: hypertension, CHD, stroke, emphysema, asthma, cancer, diabetes, smoking status, alcohol consumption status, and BMI category. These indicators are globally recognized, measurable, highly prevalent, and commonly associated with overall health status, covering six aspects: cardiovascular, metabolic, respiratory, oncological, lifestyle behavioral, and physical health. They have also been frequently used in chronic disease surveillance and population health research. Compared with individual indicators, composite HRPs may better summarize co-occurring health-risk patterns related to multimorbidity, lifestyle, and metabolic status.

Previous studies have suggested bidirectional associations between chronic health conditions and negative emotions. Negative emotions have been associated with dysregulation of the hypothalamic–pituitary–adrenal axis, autonomic nervous system activity, and chronic inflammation, whereas poor health status may contribute to psychological distress10. A previous meta-analysis reported reciprocal associations between myocardial infarction and depression, with each condition being associated with a higher likelihood of the other over time11. Individuals with asthma, chronic bronchitis, or emphysema have also been reported to experience depressive symptoms more frequently12. In addition, psychological distress has been associated with incident arthritis, cardiovascular disease, and COPD in a dose-response pattern13. Cancer and cancer-related treatment may adversely affect mental health14. Dysregulation of insulin signaling in type 2 diabetes has been linked to altered 5-hydroxytryptamine signaling and depressive symptoms, whereas depression has also been associated with hyperglycemia and poor glycemic control15. Furthermore, healthy lifestyle patterns have been associated with a lower likelihood of post-stroke depression16.

LCA for HRPs among adults with ATL

LCA is a probability-based approach for identifying latent classes from observed indicators17. In this study, LCA was used to identify descriptive, data-driven HRPs using baseline health-risk indicators rather than to define reproducible biological or clinical subtypes. LCA was applied to identify mutually exclusive latent classes. LCA was conducted by specifying to 1–5 classes and likelihood-based tests for model selection. The most appropriate number of latent classes depended on the Bayesian Information Criterion (BIC), entropy, and comparatively balanced distribution of each class. A lower BIC value was favored, whereas a high entropy value indicated better class separation, with a value close to 1 being optimal. After confirming the number of classes, individuals were assigned to a class based on their highest posterior probability. The number of classes was considered in relation to the guidance to avoid classes with fewer than 10% of the total sample. HRPs represent major components of physical health burden and are closely associated with negative emotions. To derive heterogeneous HRPs in the participants, we used LCA to classify adults with ATL in terms of baseline health-risk indicators. LCA was performed without incorporating survey weights. Therefore, the identified HRPs should be interpreted as descriptive, data-driven health-risk profiles rather than nationally representative biological or clinical subtypes. Socioeconomic status, physical activity, and other psychosocial factors were not included in the LCA model because the primary goal was to characterize physical health status subgroups rather than broader social or behavioral profiles.

Frequencies of negative emotion, nervousness, restlessness, hopelessness, effortfulness and worthlessness

Information regarding the negative emotion frequency was reported by the participants, by answering the following questions: (1) During the past 30 days, how often did you feel nervous? (2) During the past 30 days, how often did you feel restless or fidgety? (3) During the past 30 days, how often did you feel hopeless? (4) During the past 30 days, how often did you feel that everything was an effort? (5) During the past 30 days, how often did you feel worthless? The answer choices included a: none of the time, b: a little of the time, c: some of the time, d: most of the time, e: all of the time. These five items were derived from the well-validated Kessler Psychological Distress 6 Scale (K6)18. In the present study, this item combination was used as a brief measure of negative emotions and has not been validated as an independent standalone assessment tool. There is no universally accepted standard clinical threshold for defining the frequent status of this indicator. In this study, binary classification was performed according to the general clinical conventional definition of frequent occurrence. The binary indicator represented frequent recent psychological distress. This brief screening measure reflects population-level frequent distress rather than a validated clinical scale.

Choices a, b or c were coded as 0 (occasional), and choices d or e were coded as 1 (frequent) for the binary status of nervousness, restlessness, hopelessness, effortfulness, and worthlessness. The combined binary index of negative emotion frequency was defined as 0 (occasional negative emotion (ONE)) for all the codes of nervousness, restlessness, hopelessness, effortfulness, and worthlessness was 0, and 1 (FNE) for any of the codes of nervousness, restlessness, hopelessness, effortfulness and worthlessness was 1. In addition, a continuous negative emotion score was created by summing the coded frequencies of nervousness, restlessness, hopelessness, effortfulness, and worthlessness (a = 0, b = 1, c = 2, d = 3, and e = 4), with higher scores indicating greater frequencies of negative emotions.

Mortality data

Mortality data linking to the National Death Index were available through December 31, 2019.

All-cause and cause-specific mortality data were recorded. The international statistical classification of diseases and related health problems, tenth revision (ICD-10) was used to record the underlying cause of death. The duration of follow-up was defined as the interval between the interview year and the date of death and through December 31, 2019, for those participants who did not die before December 31, 2019. Rather than excluding participants with short follow-up, follow-up durations of less than 1 year were administratively set as 1 year to ensure stable estimation in survival analyses and avoid bias from extremely short follow-up periods.

Statistical analyses

Survey strata, clusters, and sampling weights were incorporated into descriptive and univariable analyses to account for the complex sampling design of the NHIS. Sociodemographic characteristics, baseline health-risk indicators, and FNE were described according to ATL status (with or without ATL). Weighted means and percentages were used to summarize the continuous and categorical variables, respectively. Differences between groups were evaluated using survey-weighted linear regression for continuous variables and survey-weighted χ² tests for categorical variables.

LCA was used to identify descriptive, LCA-derived health-risk profiles (HRPs) among adults with ATL based on baseline health-risk indicators. LCA was performed using the built-in latent class analysis module in EmpowerStats software and implemented using the poLCA package within the R statistical environment. LCA models with one to five classes were fitted using maximum likelihood estimation. Model selection was based on the BIC, entropy, and class interpretability. Participants were assigned to latent classes according to their highest posterior probabilities. Classification uncertainty was not explicitly incorporated into subsequent analyses. The associations between HRPs and FNE were subsequently examined.

We performed logistic regression analyses to examine the association between ATL and FNE. Threshold effect and smooth curve fitting analyses were applied to evaluate the potential nonlinear relationship between age and negative emotion scores according to ATL status. Associations of FNE with all-cause mortality across HRP groups were evaluated using Cox proportional hazards regression models. Cause-specific mortality analyses were additionally performed using Fine-.

Gray competing-risk regression models to account for competing non–cause-specific deaths. The subdistribution hazard ratios (SHRs) and 95% confidence intervals (CIs) were estimated. Median survival time during follow-up was estimated according to HRPs and FNE status.

Time-dependent ROC curve analysis, an extension of conventional ROC analysis for time-to-event outcomes, was conducted to assess whether adding FNE to the HRPs-based model was associated.

with changes in discrimination for all-cause mortality among adults with ATL. Net reclassification index (NRI), an index quantifying the improvement in classification accuracy of a new model compared with a reference model by reclassifying participants into more appropriate categories,

and integrated discrimination improvement (IDI), a metric assessing the global improvement in discrimination slope of a new model, of comparing sociodemographic characteristics + HRPs with sociodemographic characteristics + HRPs + FNE were calculated. Given the exploratory nature of these analyses and the absence of external validation or calibration assessment, these findings were not interpreted as evidence of clinical prediction utility.

PAFs associated with FNE for all-cause mortality among adults with ATL were exploratorily estimated to characterize the population-level distribution of the observed associations under modelling assumptions. Sensitivity analyses were additionally performed in subgroup populations.

Adjustment in the statistical analyses was used to evaluate the robustness of associations at different adjustment levels. Model I adjusted for basic sociodemographic and lifestyle-related variables, including age, sex, race, region, marital status, survey year, BMI, smoking, and alcohol consumption, which are well-established confounders for health-related studies. Model II further adjusted for major chronic comorbidities (hypertension, coronary heart disease, stroke, emphysema, asthma, cancer, and diabetes) to examine whether the observed associations remained after accounting for baseline comorbidities. Covariates were selected based on previous evidence supporting their associations with both negative emotion and mortality. Time-varying confounding was not considered in the present analysis.

All statistical analyses were performed using EmpowerStats software (http://www.empowerstats. com, X&Y Solutions, Inc., Boston, MA) based on the R environment. Other descriptive and regression analyses were conducted using standard procedures within EmpowerStats software. A two‑sided P < 0.05 was considered statistically significant.

Results

Participants characteristics

Among 635,580 adult participants from the NHIS between 1998 and 2018 with available data, 571,938 individuals with complete information on sociodemographic characteristics, baseline health-risk indicators, and ATL status were included in this study. The flow diagram of participant recruitment is shown in Fig. 1. The numbers of participants enrolled each year from 1998 to 2018 were 29,035, 27,473, 28,859, 29,223, 26,924, 26,369, 27,325, 27,329, 21,439, 20,738, 19,842, 25,748, 25,303, 30,775, 32,080, 31,462, 33,428, 30,641, 30,220, 24,385, and 23,340, respectively.

Fig. 1

Flowchart of participants selection and recruitment in our study.

Survey-weighted analyses showed significant differences in interview year, age, age group, sex, region, race, marital status, hypertension, CHD, stroke, emphysema, asthma, cancer, diabetes, smoking status, alcohol consumption status, BMI, BMI category, negative emotion score, and the frequencies of nervousness, restlessness, hopelessness, effortfulness, and worthlessness between adults with and without ATL (Table S1).

Association between ATL and FNE, nervousness, restlessness, hopelessness, effortfulness, and worthlessness

ATL was associated with FNE in the crude model, Model I, and Model II (OR = 1.66, 95% CI: 1.62–1.70, P < 0.0001; OR = 1.62, 95% CI: 1.58–1.67, P < 0.0001; and OR = 1.45, 95% CI: 1.41–1.49, P < 0.0001, respectively; Table 1). Similar associations were also observed between ATL and frequent nervousness, restlessness, hopelessness, effortfulness, and worthlessness (Table 1).

Table 1 Association between ATL with frequent negative emotion, nervousness, restlessness, hopelessness, effortfulness and worthlessness.

Full size table

Nonlinear association between age and negative emotion score according to ATL status

Among adults with ATL, age was positively associated with the negative emotion score before the age of 46 years, after adjustment for sex, race, region, marital status, survey year, and baseline health-risk indicators (coefficient: 0.03, 95% CI: 0.02–0.04, p < 0.0001, Table 2). The estimated threshold value for the negative emotion score was 4.11 (95% CI: 4.03–4.18, Table 2) and 2.24 (95% CI: 2.23–2.26, Table 2) among adults with and without ATL, respectively. Figure S1 showed the nonlinear association between age and the negative emotion score according to ATL status.

Table 2 Threshold effect analysis of age and negative emotion score according to ATL status.

Full size table

Characteristics of the identified HRPs

The two-class model showed a higher Bayesian Information Criterion (BIC) value and lower entropy (BIC: 672,169.2; entropy: 0.503; Table S2). In both the four-class and five-class models, at least one subgroup accounted for fewer than 10% of participants (Table S2). The three-class model demonstrated a comparatively lower BIC value and a more balanced class distribution (BIC: 664,720.9; entropy: 0.611; Table S2). Therefore, the three-class model was selected for subsequent analyses.

The three descriptive, LCA-derived HRPs were characterized by relatively low, moderate, and high levels of multimorbidity and adverse health behaviors, respectively. These profiles reflected descriptive patterns of co-occurring health risks rather than reproducible biological or clinical subtypes. Accordingly, the three identified profiles were designated as the low-risk HRP, intermediate-risk HRP, and high-risk HRP. Compared with the other profiles, the low-risk HRP showed lower proportions of chronic diseases, whereas the high-risk HRP showed higher proportions of chronic diseases and lower proportions of never-smokers and non-drinkers (Table 3).

Table 3 Baseline characteristics of adults with ATL based on HRPs(Weighted).

Full size table

Significant differences in interview year, age, age group, sex, region, race, marital status, hypertension, CHD, stroke, emphysema, asthma, cancer, diabetes, smoking status, alcohol consumption status, BMI, BMI category, negative emotion score, and the frequencies of nervousness, restlessness, hopelessness, effortfulness, and worthlessness were observed across HRP groups (Table 3). Figures S2-S4 illustrated the distributions of HRPs, negative emotion score, and FNE prevalence across subgroups.

Association between HRPs and FNE, nervousness, restlessness, hopelessness, effortfulness, and worthlessness

Compared with the low-risk HRP, the intermediate- and high-risk HRPs were associated with higher odds of FNE, nervousness, restlessness, hopelessness, effortfulness, and worthlessness after adjustment for age group, sex, race, region, marital status, and survey year (Table S3).

Cox regression analyses of the associations of HRPs and FNE with all-cause mortality

Compared with the low-risk HRP, the intermediate- and high-risk HRPs were associated with higher hazards of all-cause mortality (HR = 1.09, 95% CI: 1.05–1.13; and HR = 1.55, 95% CI: 1.50–1.60, respectively; Table S4). FNE was also associated with higher all-cause mortality (HR = 1.38, 95% CI: 1.33–1.43; Table S4).

Associations of negative emotion indicators with all‑cause mortality across HRP groups

Compared with the low-risk HRP + ONE group, the low-risk HRP + FNE group was associated with higher all-cause mortality during the median follow-up period of 7 years (HR = 1.33, 95% CI: 1.23–1.43; Model I; Table S5). Similarly, compared with the corresponding ONE groups, the intermediate-risk HRP + FNE group (HR = 1.35, 95% CI: 1.25–1.45) and the high-risk HRP + FNE group (HR = 1.40, 95% CI: 1.33–1.47) were also associated with higher all-cause mortality during follow-up (Model I; Table S5). Similar associations were observed for frequent nervousness, restlessness, hopelessness, effortfulness, and worthlessness (Table S5).

In addition, adults with both FNE and high-risk HRP showed a shorter median survival time during follow-up (10 years) compared with the other subgroups (Table S6).

Figure 2 illustrated the nonlinear association between the negative emotion score and survival probability among adults with ATL during the median follow-up period of 7 years. Figures S5–S7 presented the Kaplan-Meier survival curves according to FNE status, HRP groups, and their combinations.

Fig. 2

Relationship between negative emotion score and survival rate in the follow-up. time of 7 years.

Associations of FNE with cause‑specific mortality across HRP groups

Fine–Gray competing-risk regression analyses showed that, compared with the low-risk HRP + ONE group, the low-risk HRP + FNE group had a higher subdistribution hazard of cancer mortality during the median follow-up period of 7 years (Table S7). Similarly, compared with the corresponding ONE groups, the intermediate-risk HRP + FNE and high-risk HRP + FNE groups demonstrated higher subdistribution hazards of cardiovascular disease and cancer mortality during follow-up (Table S7). These findings indicate that cause-specific mortality patterns varied across HRP groups.

Exploratory time-dependent ROC analysis of models for all-cause mortality across different follow-up periods

Compared with the model including sociodemographic characteristics and HRPs, the model additionally including negative emotion frequency showed slightly higher AUC and C-index values across follow-up periods of 3, 7, 12, and 21 years (Table S8). Similarly, positive IDI and continuous NRI values were observed when comparing the two models across all follow-up periods (IDI: 0.004, 0.005, 0.004, and 0.004; continuous NRI: 0.045, 0.030, 0.028, and 0.026, respectively), suggesting only modest incremental discriminatory and reclassification information after adding negative emotion frequency.

The addition of negative emotion frequency was associated with a small improvement in population-level risk reclassification, with only approximately 3% of participants being reclassified more appropriately. Given the exploratory nature of these analyses, the limited incremental discrimination and reclassification improvements, and the absence of external validation or calibration assessment, these findings should not be interpreted as evidence of clinical prediction utility. Figure S8 illustrated time-dependent changes in the AUC for the two models during follow-up (Model I: sociodemographic characteristics + HRPs; Model II: sociodemographic characteristics + HRPs + negative emotion frequency).

PAFs associated with FNE for all-cause mortality among adults with ATL

Table S9 presents the exploratory PAF estimates associated with FNE for all-cause mortality among adults with ATL across different subgroups, together with their 95% confidence intervals. The estimated PAFs varied across age groups, sex, and follow-up periods. By age group, the estimates ranged from 3.20% (95% CI: 1.78%−4.61%) to 18.47% (95% CI: 3.97%−32.98%), but subgroup-specific estimates, particularly those for younger adults, should be interpreted cautiously because.

of likely statistical instability and the substantial uncertainty reflected by the 95% confidence intervals. The corresponding estimated PAFs were 6.41% (95% CI: 5.38%−7.44%) in females and 6.25% (95% CI: 5.09%−7.41%) in males. Across follow-up periods of 3, 7, 12, and 21 years, the estimated PAFs were 3.89% (95% CI: 2.82%−4.96%), 5.27% (95% CI: 4.39%−6.15%), 5.45% (95% CI: 4.64%−6.26%), and 6.38% (95% CI: 5.60%−7.15%), respectively (Figure S9). Given the observational nature of the study, these PAF estimates should be interpreted as exploratory, association-based population-level measures under modelling assumptions rather than causal effects or directly preventable fractions.

Sensitivity analyses

Subgroup analyses showed that ATL was associated with FNE, nervousness, restlessness, hopelessness, effortfulness, and worthlessness across most subgroups defined by age, sex, BMI, alcohol consumption status, and smoking status, except for associations with frequent nervousness and worthlessness among underweight participants and with effortfulness among participants aged ≥ 85 years (Tables S10-S14). Compared with the low-risk HRP, the high-risk HRP was associated with higher all-cause mortality across subgroups stratified by age, sex, region, race, and marital status (Table S4). Similarly, compared with the ONE group, the FNE group was associated with higher all-cause mortality across subgroups defined by age, sex, region, and marital status (Table S4). In sex-stratified analyses, the addition of FNE to the low-risk, intermediate-risk, and high-risk HRP groups was associated with higher all-cause mortality compared with the corresponding ONE groups in both males and females (Model I, Table S15).

Discussion

ATL represents an important public health concern across the world19. ATL may be associated with broader adverse effects on physical and psychological health20. In the present study, we investigated the associations of FNE with all-cause and cause-specific mortality among adults with ATL across HRP groups and evaluated whether adding FNE to the model, including sociodemographic characteristics and HRPs provided incremental discriminatory and reclassification information for all-cause mortality. Exploratory PAF analyses were also performed to characterize the population-level distribution of the observed associations.

We observed that ATL and higher-risk HRPs were associated with higher frequencies of negative emotions. Age was positively associated with the negative emotion score before the age of 46 years among adults with ATL. Consistently, stronger associations between FNE and all-cause mortality were observed among younger adults. Exploratory PAF estimates varied across age groups, with greater uncertainty among younger adults. However, these findings should be interpreted cautiously because PAF estimation relies on modelling assumptions and should not be interpreted causally, particularly in subgroups with relatively small numbers of deaths and limited statistical stability.

In addition, FNE was associated with higher all-cause mortality across all HRP groups, suggesting that the observed associations were not entirely explained by baseline HRPs. Cause-specific mortality patterns varied across HRP groups. Specifically, FNE was associated with higher cancer mortality in the low-risk HRP group, whereas both cardiovascular disease and cancer mortality were higher in the intermediate- and high-risk HRP groups. Furthermore, adding FNE to the model, including sociodemographic characteristics and HRPs was associated with only modest incremental discriminatory and reclassification information. However, given the relatively small changes in AUC, the lack of calibration assessment and external validation, and the exploratory nature of NRI and IDI analyses, these findings do not establish clinical prediction utility.

Several potential explanations may underlie the observed associations. ATL may affect facial appearance, pronunciation, eating behavior, and social interactions, which may contribute to psychological distress and negative emotions. Tooth loss has also been linked to impaired nutrition, altered sensory stimulation, and cognitive decline21,22,23. Improvements in living standards and advances in medical care have substantially increased life expectancy24. As populations age, chronic conditions such as hypertension, cardiovascular disease, and stroke become increasingly prevalent25.

In addition, chronic diseases commonly associated with ATL, including hypertension, diabetes, cardiovascular disease, and cancer, have also been associated with poorer psychological well-being26,27,28,29. Our finding that the high-risk HRP was associated with higher odds of FNE is consistent with these observations.

Negative emotions have been associated with adverse behavioral and physiological changes, including smoking, reduced physical activity, inflammation, oxidative stress, autonomic dysregulation, and impaired immune function30,31,32,33,34,35,36,37. Previous studies have also reported associations between depressive symptoms and cardiovascular disease, metabolic disorders, cancer, morbidity, and mortality31,32,33. These findings may partly explain the observed associations between FNE and mortality outcomes in adults with ATL. Nevertheless, the underlying mechanisms and temporal relationships cannot be established because this study was observational. Bidirectional relationships between physical health conditions and negative emotions are also possible.

Our study has several strengths. First, LCA based on baseline health-risk indicators allowed the identification of descriptive, LCA-derived HRPs that summarized co-occurring multimorbidity, lifestyle, and metabolic risk patterns rather than defining reproducible biological or clinical subtypes. Second, analyses across different HRP groups enabled a more comprehensive evaluation of the associations between FNE and mortality outcomes. Finally, the consistency of findings across subgroup analyses supported the relative robustness of the observed associations.

There were some limitations which merit attention. First, this was a population-based observational study with a single baseline assessment of negative emotions and covariates linked to longitudinal mortality follow-up. Reverse causality cannot be excluded because underlying physical illnesses may have contributed to elevated negative emotions at baseline. Repeated exposure measurements, lagged analyses, and formal causal inference approaches were not available. Second, chronic diseases, health behaviors, and negative emotions were self-reported and may be subject to recall bias, reporting bias, and exposure misclassification. Residual confounding remains possible because several potentially relevant variables, including psychiatric diagnoses, healthcare access, and disease severity, were unavailable. Third, the LCA-derived HRPs were descriptive and data-driven. Survey weights were not incorporated into the LCA itself, and participants were assigned to HRPs according to their maximum posterior probabilities without explicitly accounting for classification uncertainty. Therefore, these HRPs should not be interpreted as nationally representative, reproducible biological, or clinical subtypes. In addition, negative emotion was assessed using brief self-reported items referring to the previous 30 days and operationalized as a binary indicator without formal validation as an independent diagnostic instrument. Therefore, this measure may not fully capture distinct psychological constructs, symptom severity, or temporal variability in emotional status. The cutoff for frequent distress was defined for descriptive and exploratory purposes rather than according to established clinical thresholds, which may limit comparability with clinically defined conditions. Furthermore, the temporal discrepancy between short-term exposure assessment and long-term mortality follow-up may have introduced exposure misclassification, potentially influencing the observed associations. Although AUC, NRI, and IDI were used to explore the incremental information provided by FNE, the modest magnitude of these metrics, their exploratory nature and the absence of calibration assessment and external validation precluded any conclusion regarding clinical prediction utility. Finally, multiple subgroup and exploratory analyses were performed without formal correction for multiple testing, increasing the possibility of chance findings. The PAF analyses were exploratory and association-based and should not be interpreted as causal or directly preventable fractions. Subgroup-specific estimates, particularly the higher point estimate among adults aged 18–34 years, should be interpreted cautiously because of likely statistical instability and the uncertainty reflected by the 95% confidence interval. Therefore, all findings should be interpreted cautiously.

Conclusions

Taken together, our findings suggest that intermediate- and high-risk HRPs were associated with a higher likelihood of FNE, and FNE was associated with higher all-cause mortality. Cause-specific mortality patterns varied across HRP groups. Exploratory discrimination and reclassification analyses suggested that FNE provided only modest incremental information and did not establish clinical prediction utility. These findings support the epidemiological relevance of FNE among adults with ATL, but further studies with repeated psychological assessments, calibration assessment, external validation, and stronger causal designs are needed to clarify its potential clinical relevance.

来源:Nature 健康护理主题 · nature.com