Data for this study was derived from multiple cycles of the cross-sectional Continuous NHANES. The Continuous NHANES uses a complex, multistage, probability cluster design in order to yield a nationally representative survey of the health and nutritional status of the non-institutionalized civilian population in the U.S., full details on which are described in the NHANES survey methods and analytic guidelines [30]. Ethics approval was obtained from the National Center for Health Statistics Research Ethics Review Board (ERB) for NHANES 1999–2004 (Protocol #98-12), NHANES 2005–2010 (Protocol #2005-06), NHANES 2011–2016 (Protocol #2011-17) on which data for this analysis was used. Informed, written consent was obtained from all participants. This study is an analysis of NHANES publicly available anonymized data, and thus, does not require further ethical review from the York University institutional review board.

The original sample is an amalgamation of nine consecutive cycles collected in 2-year increments from 1999 through 2016 of the continuous NHANES. The initial unweighted sample size was 92,062 (Fig. 1). A final analytical sample of 13,731 was obtained for the cross-sectional analysis after listwise exclusion of individuals under 20 years of age (n = 42,550), women who were pregnant (n = 1486), those reporting (yes/no) use of mobility aids (n = 4581), cases with less than eight hours of overnight fasting (n = 21,745), cases without fasting sampling weights (n = 3970), those reporting (yes/no) any liver (n = 636) or kidney (n = 431) conditions, and missing data for covariates.

Information on demographic, lifestyle, physical measurements, and standard biochemistry profiles were collected. Standardized physical examinations, and laboratory tests, were performed in controlled environments [27]. Trained field health technicians conducted interviews at each mobile exam center (MEC), performed physical examinations (e.g., height, weight, waist circumference and blood pressure), as well as biospecimens (blood and urine) collection, which includes screening for pregnancy, storing, and shipping of biospecimens [27].

The NHANES Laboratory Method Files offer detailed description of the biological specimen collection, storage, and laboratory methods used [27]. Clinical biospecimens were often analyzed using either a Beckman Synchron LX20 prior to the 2007–2008 cycle, or a Beckman UniCel DxC 800 Synchron after the 2007–2008 cycle because of an instrumentation change in 2007 [27]. Coefficient of variation (CV) for all laboratory biomarkers were assessed using univariate analysis, and deemed to be within acceptable range (< 10%) [27, 31, 32].

Serum ALT and ALP activity were determined using a kinetic rate method, while an enzymatic rate method was used to determine AST and GGT activity [27]. An enzymatic conductivity rate method was also used to determine the concentration of blood urea nitrogen (BUN) [27]. Serum creatinine (CR) concentration was determined using a Jaffé rate method (kinetic alkaline picrate); however, CR from the periods 1999–2000 and 2005–2006 were determined to have bias due to a lack of standardized methods, or calibration [26]. In order to adjust CR for those cycles, a correction factor was recommended [27].

Glucose concentration was determined using a hexokinase-mediated reaction method, while a timed-endpoint method was used to determine the concentration of serum triglycerides [27]. Despite changes in the lab site, method, and instrumentation to measure triglycerides across cycles, no adjustments were deemed necessary as a result of these changes.

High density lipoprotein (HDL) cholesterol was analyzed in 1999–2002 primarily by heparin manganese precipitation or by direct HDL immunoassay depending on the participant age and amount of specimen [27]. Beginning in 2003, all specimens were analyzed using the direct HDL cholesterol immunoassay method, and in 2007–2008, a new laboratory performed the direct HDL cholesterol method on a different instrument [27]. HDL cholesterol values for 1999–2000, 2001–2002 and 2005–2006 were therefore adjusted to account for changes in instrumentation and methods using the CDC correction factor set out by Solomon Park Research Laboratories (Kirkland, WA) quality controls correction factor [27]. With the exception of HDL values for 1999–2002, which could not be evaluated, all laboratories were within the 5% maximum allowable bias for HDL [27].

Urinary albumin (UA) was measured using a solid-phase fluorescent immunoassay, which is a non-competitive, double-antibody method [27]. A Beckman Synchron CX3 Clinical Analyzer Jaffé rate reaction method was used to measure urinary creatinine (UCR) by enzymatic method prior to 2007 [27]. Due to an instrument change in 2007–2008, UCR was measured using Roche/Hitachi Modular P Chemistry Analyzer [27]. In accordance with recommendations, a piecewise regression approach was used to adjust UCR prior to 2007 to be comparable with UCR from 2007 onwards [27].

The outcomes for this study were all-cause and CHD mortality. The public-use versions of the LMF include a limited set of mortality variables for adult participants only after files were processed to reduce the participant disclosure risk. The public-use LMF provides mortality follow-up data from the date of survey participation through the end of the mortality period—December 31, 2015 (sample size for this analysis, n = 10,604). Data collected from the National Center for Health Statistics (NCHS) has linked several population surveys with death certificate records from the National Death Index (NDI) [35]. All-cause mortality included all known as well as unknown causes, while CHD mortality was defined as diseases of the heart using international classification of diseases coding (ICD-9 and ICD-10), including: diseases of heart (I00–I09, I11, I13, I20–I51), acute rheumatic fever and chronic rheumatic heart diseases (I00–I09), hypertensive heart disease (I11), hypertensive heart and renal disease (I13), ischemic heart diseases (I20–I25), acute myocardial infarction (I21–I22), other acute ischemic heart diseases (I24), other forms of chronic ischemic heart disease (I20, I25), and atherosclerotic cardiovascular disease (I25.0), all other forms of chronic ischemic heart disease (I20, I25.1–I25.9), and other heart diseases (I26–I51) [35].

Screening for MetS included the presence of three or more of the following five risk factors: high triglycerides (≥ 150 mg/dl); reduced high-density lipoprotein cholesterol (≤ 40 mg/dl [males] and ≤ 50 mg/dl [females]); high fasting glucose (≥ 100 mg/dl or diabetes medications); high blood pressure (systolic ≥ 130 mmHg or diastolic ≥ 85 mmHg, or hypertensive medications), and; high waist circumference (≥ 102 cm [males] and ≥ 88 cm [females]) [1, 4]. Comorbid conditions (CVD, diabetes, and hypertension) were determined using self-reports of medical conditions informed by a doctor or health professional [27].

For serum biomarkers, the Johns Hopkins Hospital’s Harriet Lane Handbook was consulted for the general population or sex-specific reference ranges [33]. For urinary biomarkers, the CDC reference ranges were consulted [27]. Biomarkers were selected and computed as ratios on the basis of previous research that established a link with insulin resistance or MetS [1, 4, 6, 15, 23]. Because biomarkers may vary by clinical laboratory settings, instrumentation, or method, a 90th percentile cut-off level was deemed as an acceptable “high-normal” to have consistency between biomarkers and sufficient group sizes for the pooled data [34]; survey cycles were pooled for the purpose of the current analysis. Elevated biomarker ratios were therefore categorized using the 90th percentile cut-offs for the pooled data as follows: BUN-CR (≥ 21.21 mg/dL), GGT-ALP (≥ 0.71 U/L), and AST-ALT (≥ 1.47 U/L). A urinary albumin to creatinine ratio (UACR) was determined by NHANES laboratory protocol [27]. Established threshold for low UACR (normoalbuminuria (< 30 mg/g)) and elevated UACR (albuminuria (≥ 30 mg/g)), were used [1, 4].

Participants were subsequently cross-classified into four discrete groups: (1) non-MetS with non-elevated biomarker ratio; (2) non-MetS with elevated biomarker ratio; (3) MetS with non-elevated biomarker ratio, and; (4) MetS with elevated biomarker ratio.

All analyses were adjusted for age, sex, ethnicity (Mexican American, other Hispanic, non-Hispanic White, non-Hispanic Black, Other/Multi-Ethnic), education, income status (a ratio of household income to the related poverty cut-off level), alcohol status (“had at least 12 drinks in a year”), lifetime smoking status (“had at least 100 cigarettes in life”), self-reported general health, health insurance coverage, healthcare utilization (frequency of hospital visits), physical activity (PA), and survey collection year [27]. Weekly minutes of moderate or vigorous-intensity PA was ascertained and used to categorize participants as “low” (< 150 min/week), or “high” (≥ 150 min/week), consistent with current PA guidelines [27].

Our first objective was to determine the prevalence and cross-sectional association between biomarkers ratios and MetS with prevalent CVD. In an initial step, socio-demographic, health, and behavioral characteristics of the sample were compared between those with versus without MetS using t-tests and chi-squared analysis, as appropriate. Biomarker ratios (mean ± SEM), and the prevalence of elevated biomarker ratios, were compared between groups with and without MetS using t-tests and chi-square, respectively. Unadjusted and fully adjusted (for covariates: age, sex, ethnicity, education, income, alcohol, smoking status, self-reported general health, health insurance coverage, frequency of hospital visits, physical activity, and survey collection year) logistic regression models were used to assess the independent and joint associations of the MetS/biomarker groups with CVD in a sensitivity analysis (n = 13,731). Using these groups, the proportional hazards assumption was subsequently assessed and upheld. For the mortality analysis (n = 10,604), a final series of Cox proportional hazards regression models (semi-parametric) were developed to examine the unadjusted and fully adjusted (for covariates) relationship for each discrete category of biomarkers with the referent (i.e., no MetS nor elevated ratios, and; prevalent MetS) and all-cause or CHD mortality (HR = 1.00, referent). A hazard ratio greater than the null value (HR = 1.00) suggests a higher risk for the event (or lower risk when less than the null). Data analysis was performed with SAS software version 9.4 (Cary, NC, U.S.A.), weighted with the master survey weights to ensure national representativeness of the data. Statistical significance was set at alpha = 0.05.

Results of the unadjusted and fully adjusted (for covariates) Cox proportional hazards models of the association between all-cause mortality and co-occurrence of MetS with kidney, or liver, biomarker ratios are presented in tables (see Additional file 1). In the fully adjusted model of the analytic sample (Fig. 2), elevated biomarker ratio with prevalent MetS (elevated UACR: HR, 95% CI = 2.57, 1.99–3.33, and elevated AST-ALT: HR = 2.22, 1.61–3.07) was positively associated with all-cause mortality, compared to the referent group—that is, the group with neither elevated biomarker ratio nor MetS. Positive associations with all-cause mortality were also observed for these elevated biomarker ratios without MetS (elevated UACR: HR = 2.12, 1.65–2.72, and elevated AST-ALT: HR = 1.71, 1.35–2.18), compared to referent. By contrast, there were no associations for the following groups: those with an elevated biomarker ratio without MetS (elevated BUN-CR: HR = 1.05, 0.79–1.39; elevated GGT-ALP: HR = 1.24, 0.92–1.67), MetS without elevated biomarker ratio (normoalbuminuria: HR = 1.05, 0.89–1.24; BUN-CR: HR = 1.10, 0.93–1.31; GGT-ALP: HR = 1.12, 0.96–1.32; AST-ALT: HR = 1.17, 1.00–1.38), and MetS with elevated biomarker ratio (elevated BUN-CR: HR = 1.26, 0.96–1.64; elevated GGT-ALP: HR = 1.26, 0.95–1.67), compared to referent. In a subsample without self-reported comorbidities (CVD, diabetes, or hypertension), the relationships in the fully adjusted (for covariates) model were maintained (Table 2); in addition, there was a statistically significant association between elevated GGT-ALP with MetS (HR = 1.39, 1.02–1.90) and all-cause mortality, compared to referent.

In fully adjusted models using prevalent MetS as the referent group (Table 3) there was a positive association between elevated biomarker ratio without MetS (elevated UACR: HR = 2.03, 1.56–2.63, and elevated AST-ALT: HR = 1.46, 1.13–1.89), and elevated biomarker ratio with MetS (elevated UACR: HR = 2.46, 1.96–3.11, and elevated AST-ALT: HR = 1.89, 1.40–2.57), and all-cause mortality. No associations were observed for the following groups: neither elevated biomarker ratio nor MetS, and; elevated biomarker ratio with/without MetS (BUN-CR and GGT-ALP).

Detailed analysis of the relationship between CHD mortality and co-occurrence of MetS with kidney, or liver, biomarker ratios are presented in the additional tables. In fully adjusted models (Fig. 2), CHD mortality was positively associated with the following groups: elevated biomarker ratio with MetS (elevated UACR: HR = 1.67, 1.05–2.67, and elevated AST-ALT: HR = 2.80, 1.62–4.86), and elevated BUN-CR without MetS (HR = 2.12, 1.12–4.04), compared to the referent group with neither elevated biomarker ratio nor MetS. No associations were observed for groups with elevated biomarker ratio without MetS (elevated UACR: HR = 1.22, 0.66–2.24; GGT-ALP: HR = 1.60, 0.80–3.21; AST-ALT: HR = 1.40, 0.76–2.58), MetS without elevated biomarker ratio (normoalbuminuria: HR = 0.83, 0.54–1.26; BUN-CR: HR = 1.17, 0.77–1.77; GGT-ALP: HR = 1.11, 0.77–1.61; AST-ALT: HR = 0.88, 0.62–1.24), or MetS with elevated biomarker ratio (BUN-CR: HR = 0.81, 0.40–1.64; GGT-ALP: HR = 0.81, 0.45–1.48), compared to referent. In a subsample without self-reported comorbidities (CVD, diabetes, or hypertension), only MetS with elevated AST-ALT (HR = 3.11, 1.49–6.49) was positively associated with CHD mortality (Table 4).

In a final set of models (unadjusted and adjusted for covariates), we explored the contribution of MetS to CHD mortality using prevalent MetS as the referent group (Table 5) and found a positive association between CHD mortality and elevated biomarker ratio with MetS (elevated UACR: HR = 2.03, 1.28–3.22, and elevated AST-ALT: HR = 3.20, 1.86–5.51). No other subgroups were related to CHD mortality.

The extent of missing covariate data was evaluated in a sample (n = 49,512) without individuals < 20 years, and the analytic sample (n = 13,731). Overall, no differences in socio-demographics, behavioral factors, or healthcare utilization and insurance coverage were found between the full and analytic sample.

Similar to the findings from this study, the presence of elevated kidney or liver biomarkers has previously been found to be associated with a higher risk of all-cause [18, 20, 22, 34], or cardiovascular mortality [17, 19, 24, 25]. Furthermore, a UACR or BUN-CR test enables the prediction of renal dysfunction or nephropathy [27, 42], including the evaluation of CVD events [8, 9, 18‐21, 38]. In terms of the surrogates for liver dysfunction, there is no single biomarker that can definitively diagnose or distinguish diseases of the liver [5, 6, 22, 23, 43]. Currently, GGT on its own has limited specificity [6, 22], and is known to be sensitive to variations in alcohol intake [42]; however, GGT-ALP could help to ascertain whether the source of a disease could be due to liver or bone pathology [22].

Routine measurements of AST and ALT support the detection of liver dysfunction [5, 6, 44]. ALT activity, however, is specific to the liver, while AST is systemic and can be found in the brain, heart, liver, lungs, kidneys, skeletal muscles, and pancreas [42]. Although AST is a less specific indicator of hepatic injury than ALT, as a ratio—AST-ALT—they tend to elevate simultaneously in cases of hepatitis, cirrhosis, hepatobiliary disorders, or necroinflammation [5, 42, 44]. In this analysis, we observed that elevated AST-ALT with and without MetS was positively associated with all-cause mortality, and; positively associated with CHD mortality in the presence of MetS, while elevated GGT-ALP with MetS was positively associated with all-cause mortality.

Evidence in support of the use of elevated AST-ALT ratio in mortality, MetS, or its comorbid risk factors is mixed, and an area of ongoing research [22, 24, 29, 45‐47]. In MetS positive cases, the AST-ALT ratio tends to be lower, which is consistent with previous studies [15, 29]. Insulin resistance, fat infiltration of the liver, or visceral adiposity [15], and medical prescriptions [24] may explain the relationship. Studies [6, 15, 24, 25] have also reported a long-term risk for metabolic disorders with normal or abnormal ALT. Taken together, however, elevated AST-ALT alongside elevated GGT or ALP has been shown to improve the identification of severe cardiometabolic disease [5, 48, 49]. Further studies are needed to investigate the potential of these biomarkers and their appropriate reference values to predict population health risks more accurately [6, 16, 22, 24, 25, 34], and may be particularly valuable for risk stratification of obesity phenotypes [47].

The main advantages of this study were: its multicycle cross-sectional design and mortality linkage, which allowed for the determination of prevalence of elevated biomarker levels and 6.9 year average survival estimates with adjustment for a host of potential confounders. The main limitations of this study were the single baseline measure of analytes and that directionality and causality could not be determined for the inter-relationship between biomarkers and MetS. Despite a large representative sample, our analyses were limited to thresholds of biomarker ratios, rather than a full examination of dose–response patterns.