Multiple cardiometabolic diseases enhance the adverse effects of hypoalbuminemia on mortality among centenarians in China: a cohort study

Study participants were recruited from the CHCCS Project, a longevity population-based study. A detailed description of study design framework has previously been published [26]. This prospective cohort study included 1,002 eligible centenarians aged more than 100 years old residing in 16 cities and counties of Hainan province from January 2014 through to December 2016. Participants were subsequently contacted until 31 May 2021 to obtain follow-up information including death for any cause.

This study was approved by the Biomedical Ethics Committee of Chinese PLA general hospital (301hn11-206-01), and was performed in accordance with the Declaration of Helsinki. All participants gave written informed consent.

We had a face-to-face interview with each study subject in the participants’ residences or nearby clinics through appropriate regional dialects. Demographic and sociological characteristics, personal disease and family histories, lifestyle and disability were collected by a standard questionnaire. All the questions on the questionnaire were asked and recorded by trained investigators from Hainan Hospital affiliated to the Chinese PLA General Hospital. Anthropometric measurements and blood pressure levels were recorded during physical examination. Body mass index is calculated as weight in kilograms divided by height squared in meters (For those with marijuana, measure their length instead of height). Blood samples were collected after a 8-h overnight fasting by venous puncture and sent to the department of biochemistry, Hainan Hospital, within 4 h, to test the level of total cholesterol (TC), triglyceride (TG), high density lipoprotein cholesterol (HDL-C), low density lipoprotein cholesterol (LDL-C), fasting blood glucose (FBG), total protein, albumin, aspartate aminotransferase (AST), alanine transaminase (ALT), hemoglobin, CRP and creatinine.

Comprehensive metabolic panel examinations were determined on Cobas 8000 automatic biochemical autoanalyzer (Roche Products Ltd., Basel, Switzerland). Albumin levels was divided into quartiles: < 36.2 g/L, 36.2–38.9 g/L, 39.0–41.2 g/L, and ≥ 41.3 g/L, and further dichotomized: more than 35 g/L and less than 35 g/L. Hypoalbuminemia is defined by a serum albumin < 35 g/L [27].

In the present analysis, CMDs included diabetes, CHD, stroke and hypertension. As part of the baseline assessment, participants were asked whether a physician had diagnosed diabetes, CHD, stroke or hypertension. We defined hypertension as previous diagnosis or SBP ≥ 140 mmHg or DBP ≥ 90 mmHg [28], and diabetes as previous diagnosis or FBG ≥ 7.0 mmol/L [29]. CHD and stroke history was finally determined from their medical records output from Grade II and III hospitals. The CMDs status was categorized as CMD-free, single CMD (any one of the above CMDs) and multiple CMDs (two or more the above of CMDs).

The death information was ascertained through linkage to the death registration system of the public security department and the registration system of the Hainan Provincial Committee on Aging, and was verified by telephone. After the survival status was confirmed, the study came to an end. There was no missing follow-up until May 31, 2021.

Socio-demographic characteristics assessed included age, sex, education, ethnicity and marriage. The ethnicity of a subject is classified as either Han or minority. Marital status is divided as widowed or not (including married, divorced, or other). The education level is classified as illiteracy or not according to the total years of education. Smoking status, alcohol drinking status and disability assessment were also collected at the same time. Smoking status and alcohol drinking status are classified into two categories: current or former, and never. Dyslipidemia was defined as previous diagnosis or TC ≥ 6.22 mmol/L, or TG ≥ 2.26 mmol/L, or HDL-C < 1.04 mmol/L, or LDL-C ≥ 4.14 mmol/L [30]. Activities of daily living (ADL) function was evaluated using the Barthel index [31], which evaluates 10 different instrumental ADL with a maximum score of 100. The ADL score was categorized into two groups to reflect degrees of ADL disability: 1) ADL independent (ADL score > 40); 2) Having an ADL disability (ADL score ≤ 40). Estimated glomerular filtration rate (eGFR) was calculated by CKD epidemiology collaboration (CKD-EPI) [32].

Continuous variables, expressed as mean (± standard deviation) in case of normally distributed variables or as median (interquartile ranges), were compared using the unpaired Student’s t-test, Mann–Whitney U-test, or One-way analysis of variance, as appropriate. Categorical variables, expressed as count (percent), were compared using the chi-square test. We used multivariable adjusted restricted cubic splines to examine possible albumin threshold levels for mortality, as well as the potential non-linear association with the risk of all-cause death. Time data for the occurrence of death events were graphically presented using the Kaplan–Meier method. After verification of proportionality assumptions, the univariate and multivariable Cox proportional risk model was used to calculate the hazard ratios (HRs) and 95% confidence intervals (CIs) of follow-up all-cause mortality associated with albumin levels and hypoalbuminemia combined with different numbers of CMDs. The models were adjusted for sex, age, ethnicity, marriage, educational levels, smoking status, alcohol drinking status, dyslipidemia, diabetes, CHD, stroke, hypertension, hemoglobin and CRP. Sensitivity analyses were conducted based on the following in further analyses to test the robustness of the primary results: a. Given to some studies add dyslipidemia in the definition of CMDs [33], we performed sensitivity analyses including dyslipidemia as a component for the definition of CMDs. b. The centenarians with ADL independent may nonetheless have higher levels of total daily physical activity compared to the centenarians with ADL disability, who might be poorer in nutritional status including hypoalbuminemia. Thus, those with ADL disability were excluded to minimize the possibility of reverse causation bias by baseline health status. c. Additional adjustment for eGFR to reduce the confounding effect of renal excretion on the relationship between albumin levels and mortality. Statistical analyses were performed using SPSS Statistics version 24.0 (IBM Corporation, Armonk, NY, United States) and R version 3.6.0 (R Foundation for Statistical Computing). Two-tailed p values of < 0.05 were considered statistically significant.

A total of 1002 participants were enrolled in our analysis, with 82.0% (822) females and median aged 102 (101–104) years old, and baseline characteristics were shown in Table 1 and Additional file 1: Table S1. Among centenarians, there was a high percentage of Han ethnicity, widowed and illiterate. The proportions of participants who never smoking and alcohol drinking were all higher than 80%. The centenarians who died had lower levels of albumin and higher prevalence of hypoalbuminemia. Albumin had a normal distribution with mean (± standard deviation) values of 38.5 g/L (± 4.0 g/L) and ranged from 23.1 to 52.2 g/L. The prevalence of hypoalbuminemia for our study was 17.4%. The distribution of albumin was depicted in Additional file 1: Figure S1. In addition, the prevalence of hypertension, diabetes, CHD and stroke were 74.7%, 9.6% 4.1% and 2.2%, respectively. The participants with lower albumin levels were tend to have significantly higher CRP levels and prevalence of ADL disability, while lower BMI, blood pressure, total protein levels and hemoglobin levels (all p < 0.05).

During the median follow-up time of 49.9 months, 522 (52.1%) all-cause deaths were observed. When treating albumin as a continuous variable, the restricted cubic spline analysis showed a flat L-shape with the inflection point at about 35 g/L (Fig. 1). A strong linear relationship was observed between a gradual rise in all-cause mortality risk and declining albumin levels, when the albumin level was < 35 g/L. However, when the albumin level was above the inflection point, the risk decreased and eventually leveled-off.

Compared with the highest quartile of albumin levels, the all-cause mortality was higher in participants with lower albumin levels (Additional file 1: Figure S2), and the multivariable-adjusted HRs for mortality in the third, second, and first quartile were 1.19 (95% CI: 0.92–1.53), 1.10 (0.85–1.44) and 1.45 (1.10–1.90) respectively, with a upward trend (P_trend = 0.021) (Table 2). In the dichotomous analysis of serum albumin using 35 g/L as the cutpoint, the cumulative incidence of death was 59.8% for participants with hypoalbuminemia and 50.5% for those with normoalbuminemia of ≥ 35 g/L, with significant differences shown in Kaplan–Meier curves (Additional file 1: Figue S3). In addition, hypoalbuminemia was associated with higher risk of all-cause mortality after multivariable adjustment (HR 1.55; 95% CI: 1.22–1.97) (Table 2).

The mortality risk for hypoalbuminemia were further analyzed taking into account the number of combined CMDs (Table 3). Among participants with hypoalbuminemia, 60.3% had single CMD and 8.6% had multiple CMDs. Compared to those with albumin levels ≥ 35 g/L, the cumulative mortality for participants with hypoalbuminemia combined with single CMD and multiple CMDs were noticeably higher (Fig. 2). The adjusted HRs were 1.45 (95%CI 1.08–1.93) and 2.15 (95%CI 1.14–4.07) respectively, and the increasing trend was significant (P_trend = 0.001). However, the number of CMDs itself was not related to a significantly increased risk of all-cause mortality (Additional file 1: Figure S4). Considering that a large proportion of centenarians with dyslipidemia, we preformed the analyses again after the inclusion of dyslipidemia into CMDs. The results were found to be similar (Additional file 1: Table S2).

Moreover, similar relationships were found after excluding participants with ADL disability or additionally adjusting for eGFR in the model, although some of the results were borderline significant due to smaller sample size (Additional file 1: Table S3, S4).