Allostatic load and progression of cardio-renal multimorbidity: A UK biobank study

The UK Biobank received approval from the North West Multi-Centre Research Ethics Committee (Approval number: 21/NW/0157), the National Information Governance Board for Health and Social Care in England and Wales, and the Community Health Index Advisory Group in Scotland, with all participants providing written informed consent. This research utilized the UK Biobank Resource (Application ID: 170605). The UK Biobank is a national prospective cohort study that enrolled over 500,000 volunteer participants, aged 37–73, from England, Wales, and Scotland between March 2006 and October 2010, with multiple follow-ups conducted subsequently. A detailed description of the cohort has been documented elsewhere [13]. At the baseline, participants were requested to furnish socio-demographic details, behavioral characteristics, and health-related information. Blood samples were gathered for genotyping and biochemistry tests. Among the 502,175 participants with available data in the current study, those with CVD or CKD at baseline (diagnosed or self-reported, n = 64,697) were excluded. We then excluded participants with missing data on select covariates (N = 40,551). Finally, 396,927 participants were involved in the current study (S1 Fig in S1 File).

Incident CVD and CKD, as well as relevant mortality, were ascertained from self-reported information, primary care data and hospital admission data. We utilized the corresponding International Classification of Diseases, 10th Revision (ICD10) or the Office of Population Censuses and Surveys Classification of Interventions and Procedures, version 4 (OPCS-4) codes to identify relevant diagnoses (S1 Table in S1 File). CVD was defined codes pertaining to coronary heart disease, atrial fibrillation, heart failure, peripheral artery disease, and stroke [14]. FCRD refereed to the first occurrence of any CVD or CKD during the follow-up period, representing the initial onset of cardio-renal impairment. CRM was defined as the coexistence of two CRDs after the initial FCRD, capturing the progression from single-organ to multi-organ involvement. These definitions reflect the clinically recognized trajectory from early cardio-renal dysfunction to multimorbidity, allowing systematic investigation of disease transitions.

AL was evaluated based on 10 biomarkers that signify three physiological systems: metabolic, cardiovascular, and inflammatory, and all were assessed at the baseline [15]. In particular, the degree of metabolic dysregulation was measured through the quantification of serum glucose (mmol/L), total cholesterol (mmol/L), HDL cholesterol (mmol/L), HbA1c (mmol/mol), IGF-1 (nmol/L), waist–to-hip ratio, and body mass index (BMI, kg/m2) [16,17]. The level of inflammatory dysregulation was determined by CRP (mg/L), whereas cardiovascular dysregulation was gauged using systolic blood pressure (SBP; mmHg) and diastolic blood pressure (DBP; mmHg) [18]. Each biomarker was then dichotomized into high risk versus low risk in accordance with sex-specific quartiles [19,20]. The study established separate high-risk thresholds for male and female participants: serum glucose, total cholesterol, glycated haemoglobin, waist-to-hip ratio, BMI, C-reactive protein, insulin-like growth factor-1, systolic blood pressure, diastolic blood pressure exceeding the 75th percentile for their respective gender, or high-density lipoprotein cholesterol below the 25th percentile for their respective gender, were classified as high-risk for that indicator. Using these gender-specific cut-off values, participants received 1 point for each high-risk indicator met and 0 points for low-risk indicators. The detailed were listed in S2 Table in S1 File. The AL score was computed by summing the 10 dichotomous scores for each of the 10 markers, yielding a score ranging from 0 to 10, where higher scores denoted a more pronounced physical dysregulation. Subsequently, a 3-group variable was considered in line with previous studies, with low (0–2), mid (3–4), and high (5–10) AL [21].

The covariates comprised age (continuous, years), ethnicity (white or other), educational level (College or University, A levels/AS level or equivalent, O levels/GCSE or equivalent, CSEs or equivalent, NVQ/HND/HNC or equivalent, other professional qualifications), socioeconomic standing (measured by Townsend deprivation index, TDI), BMI (categorized as < 18.5 kg/m², 18.5–24.9 kg/m², 25.0–29.9 kg/m², or ≥30 kg/m²), smoking status (never, previous, current), alcohol intake status (never, previous, current), dietary patterns (categorized as a healthy diet with ≥4 scores or an unhealthy diet with < 4 scores in accordance with recent guidelines for optimal cardiovascular health) [22].

Categorical variables were presented by counts (percentages), while continuous variables were depicted using either the mean (standard deviation, SD) or the median (interquartile range, IQR). The survival time for each participant was computed as the period from the baseline to the date of the incident, death, or censoring (31 October 2020), whichever came first. Two multivariable-adjusted models were established: Model 1 was adjusted for age, and ethnicity; Model 2 was adjusted for age, ethnicity, BMI, education level, TDI, dietary patterns, smoking status, and alcohol intake status.

We initially employed the Cox regression model to investigate the correlations of AL with FCRD, CRM, and death respectively. Schoenfeld residuals detected no significant violation of the proportional hazard assumptions. we then utilized multistate models to assess the association between AL and longitudinal progression from healthy to FCRD, then to CRM, and ultimately to death. There were five transitions among the three states: (A) health to FCRD, (B) health to death from a disease other than CVD, and CKD, (C) FCRD to CRM, (D) FCRD to death from any cause and (E) CRM to death from any cause. For those participants who entered different states on the same date, we determined the entry date of the theoretical prior state as the entry date of the latter state minus 0.5 days. Based on transition pattern A, we further contemplated the specific types of FCRD in disease progression. Seven transitions between five states were considered: (A) health to CVD, (B) health to CKD, (C) health to death from a disease other than CVD, and CKD, (D) CVD to CRM, (E) CKD to CRM, (F) CVD to death from any cause, (G) CVD to death from any cause and (H) CRM to death from any cause. Additionally, we further explored the potential mediating effects of WBC count and NEUT count in the association of AL with FCRD, CRM, and all-cause death using counterfactual mediation analysis.

We further conducted a series of sensitivity analyses to ensure the robustness of our results. First, we performed stratified analyses based on age, TDI, dietary patterns, smoking status and alcohol intake. Second, we repeated the analyses with complete data, employed random forest imputation for missing data, generated distinct imputed datasets (m = 5) to account for uncertainty, and set a random seed (seed = 500) for reproducibility. Third, cases of cardiovascular disease, chronic kidney disease and death occurring within two years were excluded to minimise potential reverse causation. Fourth, we replicated the analysis after excluding participants with cancer at baseline to alleviate the impact of a typically shortened lifespan linked to cancer. Fifth, we duplicated the analysis using age as the time scale to account for the influence of age on survival time. Sixth, for those participants who entered diverse states on the same day, we calculated the entry date of the prior state by employing different time intervals (30, 180, and 360 days) to assess the influence of these intervals on the outcomes. Seventh, we re-ran the analyses after omitting the cardiovascular domain from the AL when evaluating CVD risk to avoid potential statistical bias. Eighth, we reconstructed the AL score using alternative cut-points based on established clinical thresholds to test the stability of our findings across different definitions [23].

All analyses were conducted using R version 4.3.0 (R Project). The multistate model was performed using the “mstate” package. The mediation analysis was conducted using the “CMAverse” package. The Missing data were imputation using the “mice” package. P < 0.05 was considered statistically significant.

Among 396,927 participants, the mean age was 55.95 (SD = 8.07) years, 169,786 (42.96%) participants were male, and 377,437 (95.09%) were white ethnicity. Participants with higher AL tended to be men and older, have higher body mass index, less vulnerable socioeconomic status (SES), more like to smoke, and favor access alcohol intake (Table 1). In addition, during a median follow-up of 13.67 (IQR = 12.93 to 14.39) years, 122,717 (30.92%) participants experienced FCRD. Of all CRDs patients, 9,569 (7.80%) developed CRM, and, afterwards, 2,342 (24.47%) died from CRM. Additionally, 13,631 (3.43%) died without experiencing CRDs (Fig 1A). In terms of specific disease transitions, 109,362 (27.55%) incident cases of CVD, and 14,678 (3.70%) incident cases of CKD (Fig 1B). The distribution of data variables was shown in S2 Fig in S1 File.

Cox analysis revealed that a high of AL was associated with higher odds of FCRD, CRM and death (Fig 2). The multistate analysis further indicated that a high AL was associated with the elevated risk of all transitions from healthy to FCRD, to CRM and to death, particularly the transition from FCRD to CRM, followed by transition from health to FCRD. The influence of AL was broadly similar for the transitions from FCMD to Death and from CRMM to Death. Furthermore, participants with higher AL demonstrated an augmented risk of progression to CRDs and death compared to those with medium or low AL (Table 2).

We further probed into the specific types of FCRD in the dynamic progression of the disease (Transition Pattern B). It was discovered that, exception of the transition from CKD to CRM, AL exerts a remarkable influence on other transition states. Specifically, the association between AL and the transition from health to CVD as well as from CKD to CRM was the most intimate. In comparison to the transition from healthy to CVD, a higher AL was more prone to boost the risk of the transition from health to CKD. Nevertheless, the impact of AL on the transition from CVD to death was more substantial than that from CKD to death. Additionally, in contrast to individuals with medium or low AL, participants with high AL possessed a higher risk of contracting CVD and CKD (Fig 3).

The AL manifested more pronounced correlations with the advancement among participants aged 45–50 years, those with a high SES, and individuals adhering to an unhealthy diet, as opposed to older participants, those with a low SES, and individuals with a healthy diet (S3-S4 Tables inS1 File). After calculating covariates (S5 Table in S1 File, excluding outcome events that occurred during the two-year follow-up period (S6 Table inS1 File), and excluding patients with cancer at baseline (S7 Table inS1 File), there was little change in the results. Even when age was employed as the time scale and diverse time intervals were implemented for participants transitioning into different states on the same date, the robust associations between AL and the progression risk of CRDs remained conspicuous (S8-S9 Tables inS1 File). Moreover, omitting the cardiovascular domain when evaluating CVD risk (S10 Table inS1 File) and reconstructing the AL score with alternative clinical cut-points yielded concordant results, supporting the robustness of our findings (S11–S12 Tables inS1 File).