1. Introduction
For decades, low-density lipoprotein cholesterol (LDL-C) has been considered to be the major causative factor in the development of atherosclerotic cardiovascular disease (CVD) and CVD mortality [
1]. Numerous studies have robustly represented that reduction of plasma LDL-C concentration by lipid lowering agents is associated with a greater reduction in development of CVD and CVD mortality [
2,
3,
4,
5,
6,
7,
8,
9,
10].
In contrast to the enormous evidences from previous studies regarding CVD, the correlation between low plasma concentrations of LDL-C and mortality outcome is still uncertain especially in relatively healthy populations. In most randomized controlled trials or observational studies, subjects with unusually low concentration of LDL-C level have been excluded in analysis. Therefore, to date, we could not clearly find the impact of lower LDL-C on mortality outcome, especially in subjects who does not take lipid lowering agents. Moreover, some recent Japanese epidemiological studies have shown that high total cholesterol is not a risk factor for CVD and it is rather conversely associated with overall mortality [
11]. Similarly, other observational study showed that healthy individuals with low LDL-C have a significantly increased risk of both infectious diseases and cancer [
12,
13]. These studies raised an important issue whether low level of LDL-C could be related to all-cause mortality and cancer mortality in healthy populations. However, no study has evaluated the impact of LDL-C, not statin-induced decrease in LDL-C concentrations, on all-cause, cancer, and CVD mortality.
Since the effect of low concentrations of LDL-C on cancer and overall mortality remains controversial, we have investigated the associations between low levels of serum LDL-C, and cancer, all-cause mortality, and even CVD mortality in a very large, young, and well characterized, relatively healthy occupational cohort (Kangbuk Samsung health study, KSHS) during a median 5.82-year follow-up. To validate these associations, we then analyzed other dataset from a large population-based cohort study with government funding, named the Korean genome and epidemiology study (KoGES).
2. Methods
2.1. Study Population
The study population consisted of individuals who participated in a comprehensive health screening program with serum LDL-C at Kangbuk Samsung Hospital, Seoul, Korea from 2002 to 2012 (n = 396,951). The purpose of the screening program was to promote health through early detection of chronic diseases and their risk factors. Additionally, the Korean Industrial Safety and Health Law demands working individuals participate in an annual or biennial health examination. For this analysis, subjects were excluded for one or more of the following reasons: Subjects with missing data for smoking, alcohol, exercise, or lipid profiles at baseline (n = 42,020); subjects with lipid medication (n = 3667); subjects with histories of malignancy (n = 5342); subjects with mortality within 3 years after baseline (n = 649). Some of the excluded subjects had more than one of the above exclusion criteria. The total number of eligible subjects for testing associations with all-cause and CVD mortality was 347,971 (median follow up: 5.82 (IQR 2.62–8.63) years and mean (SD) follow up: 5.64 (±3.27) years). This study was approved by the Institutional Review Board of Kangbuk Samsung Hospital. Requirement for informed consent was waived as de-identified information was retrieved retrospectively.
In the validation cohort, the cohort profile of KoGES has been previously reported [
14]. The KoGES cohort was designed to investigate and assess genetic and environmental factors as correlates or determinants of the incidence of chronic diseases, (e.g., type 2 diabetes, hypertension, CVD, and cancer) in Koreans. The number of baseline subjects was 211,714. For this analysis, subjects were excluded for one or more of the following reasons: Subjects with missing data for smoking, alcohol, exercise or lipid profiles at baseline (
n = 4149); subjects with lipid medication (
n = 16,488); subjects with histories of malignancy (
n = 6578); and subjects with mortality within 3 years after baseline (
n = 1556). The total number of eligible subjects for testing associations with all-cause and CVD mortality was 183,943 (mean (SD follow up: 8.57 (±2.59) years). The percentile of women was 65.4%. At each visit, informed written consent was obtained from all participants. The study protocol was approved by the Ethics Committee of the Korean Center for Disease Control and the Institutional Review Boards of Yonsei University Wonju College of Medicine.
2.2. Data Collection
As part of the health screening program, individuals completed questionnaires related to their medical and social history and medication use. Individuals were asked about duration of education (years), frequency of exercise (none, less than once a week, at least once a week, ≥3 times per week (regular exercise)), smoking history (never, former, or current) and alcohol consumption (grams (g)/week). Trained staff also collected anthropometric measurements and vital statistics. Body weight was measured in light clothing with no shoes to the nearest 0.1 kg using a digital scale. Height was measured to the nearest 0.1 cm. Body mass index (BMI) was calculated as weight in kilograms divided by height in meters squared. Blood pressure was measured using standard mercury sphygmomanometers. Blood samples were collected after minimum 10 h of fasting and analyzed in the same core clinical laboratory. The core clinical laboratory has been accredited and participates annually in inspections and surveys by the Korean Association of Quality Assurance for Clinical Laboratories. Serum levels of total cholesterol, triglycerides, LDL-C, and high-density lipoprotein cholesterol (HDL-C) were measured using Bayer Reagent Packs (Bayer Diagnostics, Leverkusen, Germany) on an automated chemistry analyzer (Advia 1650 Autoanalyzer; Bayer Diagnostics, Leverkusen, Germany).
Deaths among participants were identified by matching the information to death records from the National Statistical Office using identification numbers assigned to subjects at birth. Causes of death were coded centrally by trained coders using the ICD-10 classification (International Classification of Diseases, 10th revision). In this study, CVD mortality was defined as ICD-10 codes I00 to I99.
2.3. Statistical Analysis
The statistical analysis was performed using STATA version 14.0 (StataCorp LP, College Station, TX, USA). Reported
p values were two-tailed, and <0.05 were considered statistically significant. The distribution of continuous variables was evaluated and transformations were conducted for nonparametric variables. We divided our subjects according to plasma LDL-C concentrations (<70, 70–99, 100–129, 130–159, ≥160 mg/dL) at baseline. Cox proportional hazards models stratified by five groups were used to estimate hazard ratios (HRs and 95% CIs for all-cause mortality, CV, and cancer mortality in each LDL-C category, compared with the LDL-C 100–129 mg/dL as the reference group). This LDL-C 100-129 mg/dL group was chosen as the reference because this group contained the mean LDL-C concentration for adults in Korea over the last 10 years (approximately 110 mg/dL) [
15]. For testing linear risk trends across LDL-C concentration groups in the regression models, we used the categories rank as a continuous variable. To minimize the influence of possible “reverse causation” (illnesses causing low LDL-C), we excluded the subjects who died with in less than 3 years after the baseline measurements. A cubic spline analysis was used to characterize non-linear, dose-response associations between LDL cholesterol levels and mortality, and to minimize residual confounding for continuous confounders [
16,
17]. We checked the proportional hazards assumption by examining graphs of estimated log (-log) survival.
p < 0.05 was considered significant.
4. Discussion
Our novel results show that low levels of LDL-C (<70 mg/dL) were associated with increased risk of CVD mortality, cancer mortality, and even all-cause mortality especially in men who were not treated with lipid lowering therapy. The finding of increased CVD mortality in men with low levels of LDL-C (<70 mg/dL) was observed in both different cohorts even though it showed a U shape. In this study, we were able to take account of multiple confounders and the young age of the cohort helped decrease the influence of potential reverse causality between clinically relevant outcomes and low levels of plasma LDL-C concentrations. Additionally, we excluded subjects who died within 3 years of follow up to avoid the possibility of reverse causality. Furthermore, to validate these associations, we then analyzed other dataset from a large population-based cohort study with government funding, named the Korean genome and epidemiology study (KoGES) which consists of community-dwellers aged ≥40 years at baseline.
We chose the third LDL-C group (i.e., LDL-C 100–129 mg/dL) as the reference group, because as indicated above this group contained the mean LDL-C concentration for the Korean population as measured in the Korean National Health and Nutrition Examination Survey during 1998 to 2010 [
15]. As we expected, the highest category of LDL-C (≥160 mg/dL or ≥130 mg/dL in KoGES) was associated with increased risk of CVD mortality. However, the lowest LDL-C concentration category (LDL-C <70 mg/dL) also showed higher risk of CVD mortality compared to the reference group.
In line with our findings, another recent study also presented that whereas low LDL-C (<70 mg/dL) was not associated with protective effects on CVD outcome, low hs-CRP appeared to be associated with reduced risk of incident CVD and CVD mortality in high risk population [
18]. These findings provide a paradoxical contradiction to the traditional LDL-C hypothesis; a lower CVD and all-cause mortality in lower LDL-C levels. It suggests the possibility that lower LDL-C concentration itself may not be a crucial factor for health outcome and other factor such as inflammatory process may have more important role in health outcome. However, considering the known strong association lowering LDL-C levels and better CV outcome, our finding indicates potential higher risk of poor health outcome in subjects who have too lower level of LDL-C although they do not take lipid lowering agents.
Associations between lower levels of LDL-C and poor health outcome have been reported in some, but not all, prior studies. Observational cohort studies have revealed that people with low total cholesterol levels (e.g., total cholesterol < 154.4 mg/dL) have increased risk of subsequent death in some cancers, respiratory diseases, and other non-medical causes than people with high baseline cholesterol levels [
4]. A recent systematic review of 19 cohort studies including more than 68,000 elderly people showed that CVD mortality was highest in the lowest LDL-C quartile group [
19]. However, these studies included participants who were taking lipid-lowering agents and who had other co-morbidities which may have influenced outcomes. Our study has excluded all subjects who were taking any lipid-lowering therapy at baseline in order to investigate the direct association between low levels of LDL-C and mortality outcomes. We demonstrated an increase in any cause of mortality outcomes in the lowest LDL-C concentration group especially in men. The finding of increased risk of mortality in men with low level of LDL-C was similar when we even excluded subjects who have history of diabetes and CVD at baseline. We additionally confirmed this phenomenon in another validation cohort. Our finding provide evidence supporting the ‘lipid paradox’, suggesting that too lower level of cholesterol concentrations do not always confer protective effects on mortality outcomes in the healthy population who does not take lipid-lowering agents.
While the exact mechanism remains to be elucidated, several possibilities could explain our findings. Firstly, a low LDL-C concentration increases susceptibility to fatal disease. Some experiments have shown that LDL-C binds to and inactivates a broad range of microorganisms and toxic products which might be a possible causal factor of CVD and cancer [
20,
21]. Furthermore, a common mechanism may operate that links low LDL-C concentration to different disease states. Links between low LDL-C and death from different diseases, only seems plausible if low LDL-C concentration is a marker for another phenomenon and to this effect although it is pure speculation, we and others have suggested that dysbiosis and altered bile acid metabolism [
22,
23,
24,
25] could provide that common link.
There are strengths and limitations of our study that should be considered in the interpretation of these controversial data. A number of 347,971 relatively young subjects (mean age 39.6 years) (57.4% men) were studied in a retrospective cohort study design over a median follow up of almost 6 years and data on cardiovascular mortality in men validated in another independent cohort. Additionally, we have excluded the data from individuals who were identified at baseline and who subsequently died during the first three years of follow up. These factors limit the possibility of reverse causality explaining our findings. However, since we excluded all subjects at baseline who were taking any lipid-lowering therapy, it is likely that subjects with extremely highest level of LDL-C have been excluded. Moreover, the weaknesses of our study design is that treatment with LDL-C lowering therapy during the period of follow up is not available, although given what is known about the benefit of statins, treatment with statins would decrease CVD and misclassification bias would operate to bias our results towards the null. Additionally, the fact that the numbers of deaths, especially cardiovascular mortality, are relatively low may attenuate the causal relationship between LDL cholesterol and mortality. We could not measure some specific lipoproteins such as small dense LDL and lipoprotein, which may explain this phenomenon. Another limitation of our study is that the sample was limited to Korean, relatively young-aged participants, and it is uncertain whether the findings are applicable to other ethnic groups.