Comparative Prognostic Performance of Nutritional, Immune-inflammatory, and Anthropometric Indices for Cardiovascular and All-cause Mortality in Older Adults with Cardiovascular Disease (NHANES 2009–2018) Authors: Hoang Le Huy Nguyen Published: 2026 ABSTRACT AbstractBackgroundCardiovascular disease (CVD) in older adults is compounded by malnutrition and chronic inflammation. We compared the long-term prognostic performance of three categories of routinely available indices: nutritional (Prognostic Nutritional Index [PNI], Geriatric Nutritional Risk Index [GNRI], Controlling Nutritional Status [CONUT]); immune-inflammatory (Systemic Immune-Inflammation Index [SII]); and anthropometric (Weight-Adjusted Waist Index [WWI]).MethodsWe analyzed 1,804 National Health and Nutrition Examination Survey 2009–2018 participants aged ≥60 years with established CVD, linked to the National Death Index. Survey-weighted Cox models with restricted cubic splines estimated associations with cardiovascular and all-cause mortality, adjusted for demographic, socioeconomic, behavioral, and comorbid covariates. Sensitivity analyses included landmark exclusion of early deaths (6, 12, 24 months), a complete-case analysis, and additional adjustment for total energy and protein intake.ResultsPNI and GNRI showed L-shaped associations with cardiovascular mortality, whereas CONUT showed a monotonic positive association. Each 1-point increase in PNI corresponded to an adjusted hazard ratio of 0.91 (95% confidence interval 0.87–0.95) for cardiovascular mortality; moderate-to-severe malnutrition by PNI carried a 4.61-fold higher hazard. PNI yielded the best model fit for cardiovascular mortality (lowest Akaike and Bayesian information criteria [AIC and BIC]); CONUT had the highest discrimination. The SII association attenuated when deaths within 24 months were excluded. The WWI predicted all-cause but not cardiovascular mortality.ConclusionThe nutritional indices (PNI, GNRI, CONUT) showed durable associations with cardiovascular mortality, whereas the SII signal was acute and the WWI predicted only all-cause mortality. These findings are hypothesis-generating; prospective external validation is required before routine clinical use.Keywords: Aged, Cardiovascular diseases, Mortality, Nutrition Assessment, Inflammation INTRODUCTION Cardiovascular disease (CVD) is the leading cause of death globally and is concentrated in older adults.1,2) Aging is accompanied by inflammaging and immunosenescence3) that, together with age-related shifts in body composition, drive the malnutrition–inflammation–atherosclerosis (MIA) syndrome.4) Because traditional cardiovascular risk algorithms were developed largely in middle-aged populations, they incompletely capture this geriatric vulnerability.5,6) Composite indices that summarize nutritional, inflammatory, and anthropometric status have therefore been proposed. Five such indices, derived from routine clinical data, are examined here: three nutritional indices (Prognostic Nutritional Index [PNI], Geriatric Nutritional Risk Index [GNRI], and Controlling Nutritional Status [CONUT] score), one immune-inflammatory marker (Systemic Immune-Inflammation Index [SII]), and one anthropometric measure (Weight-Adjusted Waist Index [WWI]). The PNI predicts long-term mortality after percutaneous coronary intervention for acute coronary syndromes.7) The GNRI, designed for geriatric populations, predicts adverse outcomes in patients at risk for heart failure (HF),8) in frail older adults after percutaneous coronary intervention,9) and in older people with declining renal function.10) The CONUT score is an independent predictor of mortality in HF11) and in hypertensive cohorts.12) The SII reflects systemic inflammation; the WWI quantifies central adiposity with less weight-related confounding than body mass index (BMI). Large-scale analyses link these indices to all-cause and cardiovascular mortality across diverse adult populations.13-17) Existing evidence is limited to single-index designs in narrow clinical cohorts, precluding direct comparisons and limiting generalizability.15,17) No study has evaluated all five indices in parallel in a nationally representative sample. We used the National Health and Nutrition Examination Survey (NHANES) to compare their prognostic performance in older adults with established CVD, with cardiovascular mortality as the primary outcome and all-cause mortality as the secondary outcome. MATERIALS AND METHODS Study Design and Participants This retrospective cohort study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines.18) Data were drawn from the NHANES, a continuous program of the National Center for Health Statistics (NCHS) that uses a multistage probability sampling design to generate nationally representative estimates for the non-institutionalized U.S. population through in-home interviews, physical examinations, and laboratory testing at a mobile examination center (MEC).19) Five NHANES cycles (2009–2018) were pooled, yielding a source population of 49,693. Sequential exclusions removed participants aged <60 years (n=39,937); those without a self-reported physician diagnosis of established CVD (congestive HF, coronary heart disease, angina pectoris, myocardial infarction, or stroke; n=7,383); those lacking National Death Index linkage (n=4); those with hematologic abnormalities (white blood cells >30,000/µL, lymphocytes >10,000/µL, or platelets >600,000/µL; n=14); and those with missing data for any of the five indices (n=551). The final analytic cohort comprised 1,804 participants (Fig. 1). Derivation of the analytic cohort. Of 49,693 NHANES 2009–2018 participants, sequential exclusions left 1,804 adults aged ≥60 years with established cardiovascular disease for analysis. Exclusions: age <60 years (n=39,937); no established cardiovascular disease, defined as heart failure, coronary heart disease, angina, myocardial infarction, or stroke (n=7,383); missing mortality linkage (n=4); hematologic outliers, defined as white-cell count >30,000/µL, lymphocyte count >10,000/µL, or platelet count >600,000/µL (n=14); and missing data on any of the five indices (n=551). Over follow-up, 217 participants died of cardiovascular causes, 288 of non-cardiovascular causes (total all-cause deaths, 505), and 1,299 remained alive at the end of follow-up. NHANES, National Health and Nutrition Examination Survey. The NCHS Research Ethics Review Board approved the NHANES protocol, and all participants provided written informed consent; this secondary analysis of de-identified public data required no further review. Data Collection and Measurements Predictors Five candidate indices were evaluated across three categories: three nutritional indices (PNI, GNRI, CONUT score), one immune-inflammatory marker (SII), and one anthropometric index (WWI). Nutritional indices The PNI was calculated as 10 × albumin (g/dL) + 0.005 × total lymphocyte count (TLC,cells /mm³)20,21) and categorized as normal (>50), mild (45–50), or moderate-to-severe (<45). The GNRI was calculated as 1.489 × albumin (g/L) + 41.7 × (current/ideal weight), with ideal weight from the Lorentz formula and the weight ratio capped at 122); participants were classified as no risk (≥98) or at risk (<98), combining the originally proposed low- and moderate-to-severe-risk strata to ensure adequate group sizes. The CONUT score summed deficits in serum albumin, TLC, and total cholesterol23) and was dichotomized as normal (0–1) or abnormal (≥2), collapsing the original three-level scheme owing to few decedents in the highest stratum. Immune-inflammatory marker The SII was calculated as platelet count × neutrophil count/lymphocyte count24) and natural log-transformed for continuous analyses because of its right-skewed distribution. In the absence of validated cut-offs in older adults with CVD, the SII was categorized by cohort-specific tertiles25): low (<384.0), middle (384.0–608.5), and high (≥608.5). Anthropometric index The WWI was calculated as waist circumference (cm) divided by the square root of body weight (kg)26,27) and similarly categorized by cohort-specific tertiles: low (<11.33), middle (11.33–12.00), and high (≥12.00). Outcomes The primary outcome was cardiovascular mortality (International Classification of Diseases, 10th Revision [ICD-10] underlying-cause codes I00–I99); the secondary outcome was all-cause mortality. Vital status was ascertained by probabilistic linkage to the National Death Index through December 31, 2019.28) Covariates Covariates were obtained from in-home interviews and MEC examinations. Age was categorized as 60–69, 70–79, or ≥80 years; sex (as recorded by NHANES) and race/ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic, other) were self-reported. Education was dichotomized as below high school versus high school or above. The poverty-income ratio (PIR) was categorized as low (<1.30), middle (1.30–4.99), or high (≥5.00). Smoking status was categorized as never (<100 lifetime cigarettes), former, or current; alcohol use as nondrinker or current drinker (any use in the past 12 months). BMI was analyzed continuously and as underweight (<18.5 kg/m²), normal (18.5–24.9), overweight (25.0–29.9), or obese (≥30.0). Hypertension, diabetes, and chronic kidney disease (CKD) were defined by self-reported diagnosis, medication use, or clinical thresholds (blood pressure ≥140/90 mmHg, hemoglobin A1c ≥6.5%, estimated glomerular filtration rate <60 mL/min/1.73 m², or albumin-to-creatinine ratio ≥30 mg/g). Serum albumin, total cholesterol, and total lymphocyte count were analyzed as continuous variables. Quality Control NHANES applies standardized operating procedures and quality-assurance protocols across all cycles, including technician training, laboratory calibration, and interlaboratory reliability checks.29) Prior to analysis, NHANES codes for "refused" (7, 77, 777) and "don’t know" (9, 99, 999) were recoded as missing, and skip-pattern logic was reviewed against the component codebooks. Statistical Analysis Analyses were performed in R 4.6.0 (R Foundation for Statistical Computing, Vienna, Austria) with the survey package, incorporating sample weights, masked variance units, and subpopulation analysis per NCHS Analytic Guidelines.29) The 2-year MEC examination weights were divided by five and summed to construct a 10-year analytic weight. Continuous variables are reported as weighted means (standard error) for approximately normal distributions or medians (interquartile range) for skewed distributions, assessed from weighted histograms29); skewed variables were log-transformed. Categorical variables are presented as unweighted counts with weighted percentages (standard error). Participants with and without cardiovascular death were compared using survey-weighted linear regression for continuous variables and survey-weighted logistic regression for categorical variables; for variables with more than two categories, the reported p-value was the Bonferroni-adjusted minimum across non-reference Wald contrasts, capped at 1.000. Estimates not meeting NCHS Data Presentation Standards for Proportions29) are flagged in table footnotes. Missing data were handled by multiple imputation with chained equations (20 datasets), incorporating the survey design and including all analytic covariates, mortality status, the Nelson-Aalen cumulative hazard, and the continuous indices as auxiliary variables. Estimates were pooled by Rubin’s rules. All records with a positive sample weight were retained per NCHS guidance.29) Statistical significance was set at two-sided p<0.05; effect estimates are reported with 95% confidence intervals (CIs). Associations between each index and mortality were estimated with survey-weighted Cox proportional hazards models; results are reported as adjusted hazard ratios (aHRs) with 95% CIs. Follow-up time was measured from the MEC examination date to death or December 31, 2019, whichever occurred first. Three nested models were prespecified: Model 1, unadjusted; Model 2, adjusted for age category, sex, race/ethnicity, education, and PIR; Model 3, Model 2 plus smoking, alcohol use, hypertension, diabetes, and CKD. Model 3 was used for primary analyses, including restricted cubic splines with four knots (three for CONUT, given its discrete distribution) to assess nonlinear dose–response. Proportional hazards were verified by scaled Schoenfeld residuals. Discrimination was compared across fully adjusted models using Harrell’s C-statistic pooled across imputed datasets; model fit was compared using the Akaike (AIC) and Bayesian (BIC) information criteria. Three prespecified sensitivity analyses were performed. First, to assess reverse causality, landmark analyses were performed in which fully adjusted models were refit after sequentially excluding early deaths, operationally defined as deaths occurring within 6, 12, or 24 months of the MEC examination (Supplementary Tables S1 and S2). Second, multiply imputed results were compared with a complete-case analysis, with the complete-case subset applied to the full survey design via subpopulation analysis (Supplementary Table S3). Third, to assess residual dietary confounding, Model 3 was further adjusted for total energy and protein intake from the NHANES Day-1 24-hour dietary recall (United States Department of Agriculture Automated Multiple-Pass Method),29) using the 10-year Day-1 dietary weight and restricted to participants with a complete and reliable Day-1 recall; dietary intake was not imputed (Supplementary Table S4). RESULTS Study Population Of the 49,693 participants enrolled in NHANES 2009–2018, 1,804 older adults with established CVD met the eligibility criteria; stepwise exclusions and outcome assignment are shown in Fig. 1. The analytic cohort comprised 217 participants who died of cardiovascular causes and 1,587 participants with no cardiovascular death, including 1,299 alive at the end of follow-up and 288 who died of non-cardiovascular causes. Participants were predominantly male and Non-Hispanic White, with a high prevalence of hypertension and obesity. Participants with cardiovascular death were older than those without cardiovascular death and more frequently had HF and CKD; they had lower mean PNI, lower median lymphocyte count, and higher median SII. Baseline characteristics with stratum-specific estimates and p-values are reported in Table 1. Table 1.Baseline characteristics of the study cohort, according to cardiovascular mortality statusCharacteristicTotal (n=1,804)Cardiovascular Death (n=217)No Cardiovascular Death (n=1,587)p-valueNutritional indices PNI50.6±0.249.0±0.450.8±0.2<0.001 GNRI102.7±0.2101.7±0.4102.8±0.20.009 CONUT0.007 Normal (0–1)1,087 (59.7)108 (49.1)979 (60.8) Abnormal (≥2)717 (40.3)109 (50.9)608 (39.2)Immune-inflammatory index SII496.1 (336.5–678.0)528.0 (410.7–804.7)493.1 (330.3–665.6)<0.001 SII tertile0.008 Low (T1)618 (33.0)56 (22.3)562 (34.1) Middle (T2)596 (34.0)75 (36.6)521 (33.7) High (T3)590 (33.0)86 (41.1)504 (32.1)Anthropometric index WWI11.68±0.0311.77±0.0711.68±0.030.205 WWI tertile1.000 Low (T1)571 (33.0)66 (32.6)505 (33.0) Middle (T2)637 (34.1)67 (29.6)570 (34.6) High (T3)596 (32.9)84 (37.8)512 (32.4)Demographics and socioeconomic Age group <0.001 60–69 y688 (41.1)44 (18.6)644 (43.5) 70–79 y648 (35.6)68 (30.3)580 (36.1) ≥80 y468 (23.4)105 (51.1)363 (20.4) Male1,086 (57.4)139 (55.3)947 (57.6)0.652 Race/ethnicity0.439 Non-Hispanic White988 (77.8)140 (80.7)848 (77.5) Non-Hispanic Black345 (8.2)40 (9.0)305 (8.1) Hispanic322 (6.6)30 (6.8)292 (6.6) Other149 (7.4)7 (3.5)142 (7.8) High school or above1,228 (78.7)141 (72.4)1,087 (79.4)0.072 PIR category0.078 <1.30 (low)369 (12.3)44 (14.1)325 (12.1) 1.30–4.99 (middle)1,101 (59.4)144 (68.3)957 (58.4) ≥5.00 (high)334 (28.3)29 (17.6)305 (29.5)Lifestyle Smoking status0.960 Never723 (40.1)93 (42.5)630 (39.8) Former786 (44.9)88 (44.1)698 (45.0) Current295 (15.0)36 (13.4)259 (15.2) Current alcohol use938 (55.2)91 (42.7)847 (56.6)0.004Anthropometry BMI (kg/m²)29.1 (25.7–33.7)28.0 (24.6–32.3)29.2 (25.8–33.8)0.043 Waist circumference (cm)105.2 (96.0–116.0)104.1 (93.4–113.3)105.4 (96.6–116.1)0.028 BMI category0.011 Underweight (<18.5 kg/m²)20 (0.9)5 (3.1)15 (0.7) Normal (18.5–24.9 kg/m²)376 (20.0)50 (23.7)326 (19.6) Overweight (25.0–29.9 kg/m²)637 (33.8)76 (34.1)561 (33.7) Obesity (≥30.0 kg/m²)771 (45.3)86 (39.1)685 (46.0)Cardiovascular phenotypes and comorbidities Heart failure536 (28.0)86 (43.7)450 (26.3)<0.001 Coronary heart disease783 (47.6)97 (50.1)686 (47.3)0.566 Angina pectoris415 (24.2)45 (22.1)370 (24.5)0.557 Myocardial infarction723 (38.6)90 (38.5)633 (38.6)0.994 Stroke594 (29.7)71 (31.0)523 (29.6)0.782 Hypertension1,494 (80.4)187 (85.8)1,307 (79.8)0.068 Diabetes mellitus789 (39.9)106 (44.9)683 (39.3)0.251 CKD945 (46.9)165 (75.3)780 (43.8)<0.001Laboratory parameters Serum albumin (g/dL)4.11±0.014.06±0.034.12±0.010.071 Total cholesterol (mg/dL)168.0 (144.0–198.0)165.0 (143.0–197.0)169.0 (144.0–199.0)0.172 Neutrophils (×10³/µL)4.3 (3.4–5.4)4.5 (3.5–5.7)4.2 (3.3–5.3)0.056 Lymphocytes (×10³/µL)1.8 (1.4–2.3)1.5 (1.2–2.0)1.8 (1.4–2.3)<0.001 Platelets (×10³/µL)204.0 (171.0–249.0)192.0 (155.0–248.0)204.0 (174.0–249.0)0.064Values are presented as mean±standard error, n (weighted %), or median (interquartile range). All summaries incorporate the 10-year mobile examination center (MEC) sample weight, primary sampling units, and strata.PNI, Prognostic Nutritional Index; GNRI, Geriatric Nutritional Risk Index; CONUT, Controlling Nutritional Status; SII, Systemic Immune-Inflammation Index; WWI, Weight-Adjusted Waist Index; PIR, poverty–income ratio; BMI, body mass index; CKD, chronic kidney disease.The analytic cohort comprised 1,804 National Health and Nutrition Examination Survey 2009–2018 participants aged ≥60 years with established cardiovascular disease.p-values are from survey-weighted linear regression for continuous variables (log-transformed when right-skewed) and survey-weighted logistic regression for categorical variables; for variables with more than two categories, p-values are Bonferroni adjusted across non-reference Wald contrasts and capped at 1.000. Baseline characteristics of the study cohort, according to cardiovascular mortality status Values are presented as mean±standard error, n (weighted %), or median (interquartile range). All summaries incorporate the 10-year mobile examination center (MEC) sample weight, primary sampling units, and strata. PNI, Prognostic Nutritional Index; GNRI, Geriatric Nutritional Risk Index; CONUT, Controlling Nutritional Status; SII, Systemic Immune-Inflammation Index; WWI, Weight-Adjusted Waist Index; PIR, poverty–income ratio; BMI, body mass index; CKD, chronic kidney disease. The analytic cohort comprised 1,804 National Health and Nutrition Examination Survey 2009–2018 participants aged ≥60 years with established cardiovascular disease. p-values are from survey-weighted linear regression for continuous variables (log-transformed when right-skewed) and survey-weighted logistic regression for categorical variables; for variables with more than two categories, p-values are Bonferroni adjusted across non-reference Wald contrasts and capped at 1.000. Nutritional Indices and Cardiovascular Mortality In survey-weighted Cox proportional hazards models, each of the three nutritional indices was independently associated with cardiovascular mortality after full adjustment, on both the continuous scale and across categorical strata (Model 3 in Table 2). Associations were inverse for PNI and GNRI and direct for CONUT score. Moderate-to-severe malnutrition by PNI was associated with more than a four-fold hazard, and at-risk GNRI with more than a two-fold hazard, relative to normal-range references (Table 2). Table 2.Associations of nutritional, immune-inflammatory, and anthropometric indices with cardiovascular mortalityIndex and comparisonModel 1Model 2Model 3aHR (95% CI)p-valueaHR (95% CI)p-valueaHR (95% CI)p-valueNutritional indices PNI, per 1-unit increase0.89 (0.85–0.93)<0.0010.91 (0.87–0.95)<0.0010.91 (0.87–0.95)<0.001 PNI, normal (>50, reference)1.00 (ref)1.00 (ref)1.00 (ref) PNI, mild risk (45–50)1.58 (1.05–2.36)0.0271.38 (0.89–2.14)0.1461.37 (0.87–2.17)0.173 PNI, moderate-to-severe risk (<45)5.04 (3.29–7.72)<0.0014.75 (3.14–7.20)<0.0014.61 (3.09–6.87)<0.001 GNRI, per 1-unit increase0.91 (0.88–0.95)<0.0010.92 (0.89–0.96)<0.0010.92 (0.89–0.96)<0.001 GNRI, no risk (≥98, reference)1.00 (ref)1.00 (ref)1.00 (ref) GNRI, at risk (<98)2.69 (1.77–4.06)<0.0012.34 (1.56–3.51)<0.0012.36 (1.58–3.54)<0.001 CONUT, per 1-point increase1.45 (1.27–1.65)<0.0011.37 (1.22–1.54)<0.0011.38 (1.22–1.55)<0.001 CONUT, normal (0–1, reference)1.00 (ref)1.00 (ref)1.00 (ref) CONUT, abnormal (≥2)1.85 (1.38–2.49)<0.0011.58 (1.19–2.10)0.0021.61 (1.19–2.17)0.002Immune-inflammatory index log(SII), per 1-unit increase1.61 (1.24–2.08)<0.0011.52 (1.19–1.94)<0.0011.45 (1.13–1.86)0.003 SII, low tertile (reference)1.00 (ref)1.00 (ref)1.00 (ref) SII, middle tertile1.57 (1.01–2.44)0.0451.54 (0.98–2.42)0.0591.58 (0.97–2.56)0.065 SII, high tertile1.88 (1.24–2.86)0.0031.85 (1.22–2.80)0.0041.71 (1.12–2.62)0.013Anthropometric index WWI, per 1-unit increase1.36 (1.05–1.76)0.0201.23 (0.94–1.62)0.1321.12 (0.86–1.46)0.387 WWI, low tertile (reference)1.00 (ref)1.00 (ref)1.00 (ref) WWI, middle tertile0.98 (0.60–1.60)0.9240.86 (0.53–1.41)0.5570.80 (0.50–1.29)0.363 WWI, high tertile1.41 (0.87–2.28)0.1591.21 (0.76–1.95)0.4201.03 (0.65–1.64)0.887PNI, Prognostic Nutritional Index; GNRI, Geriatric Nutritional Risk Index; CONUT, Controlling Nutritional Status; SII, Systemic Immune-Inflammation Index; WWI, Weight-Adjusted Waist Index; aHR, adjusted hazard ratio; CI, confidence interval.Hazard ratios were estimated with survey-weighted Cox proportional hazards models, pooled across 20 multiply imputed datasets using Rubin's rules. Model 1 was unadjusted; Model 2 was adjusted for age category, sex, race/ethnicity, education, and poverty–income ratio; Model 3 was Model 2 plus smoking status, current alcohol use, hypertension, diabetes mellitus, and chronic kidney disease. The SII was natural-log-transformed because its distribution was right-skewed; the per-unit hazard ratio is therefore on the log scale. Cohort-specific tertiles were used for the SII and WWI in the absence of validated cut-offs in older adults with cardiovascular disease. Associations of nutritional, immune-inflammatory, and anthropometric indices with cardiovascular mortality PNI, Prognostic Nutritional Index; GNRI, Geriatric Nutritional Risk Index; CONUT, Controlling Nutritional Status; SII, Systemic Immune-Inflammation Index; WWI, Weight-Adjusted Waist Index; aHR, adjusted hazard ratio; CI, confidence interval. Hazard ratios were estimated with survey-weighted Cox proportional hazards models, pooled across 20 multiply imputed datasets using Rubin's rules. Model 1 was unadjusted; Model 2 was adjusted for age category, sex, race/ethnicity, education, and poverty–income ratio; Model 3 was Model 2 plus smoking status, current alcohol use, hypertension, diabetes mellitus, and chronic kidney disease. The SII was natural-log-transformed because its distribution was right-skewed; the per-unit hazard ratio is therefore on the log scale. Cohort-specific tertiles were used for the SII and WWI in the absence of validated cut-offs in older adults with cardiovascular disease. Immune-inflammatory and Anthropometric Indices, and Cardiovascular Mortality After full adjustment, natural-log-transformed SII remained associated with cardiovascular mortality; the highest tertile carried a higher hazard than the lowest (Table 2). The WWI was not associated with cardiovascular mortality on either the continuous scale or across tertiles (Table 2). Kaplan–Meier curves separated by strata of each nutritional index and of the SII, with attenuated separation for WWI (Fig. 2). Restricted cubic-spline analyses showed L-shaped dose–response curves for PNI and GNRI, with steep hazard increases below nutritional thresholds; a monotonic positive curve for CONUT; a J-shaped curve for log-SII; and a near-flat curve for WWI (Fig. 3). The proportional-hazards assumption was satisfied for all fully adjusted models (Schoenfeld residual tests, all p>0.10). Survey-weighted Kaplan–Meier survival curves for cardiovascular and all-cause mortality, according to each index. Curves show cardiovascular mortality (Panels A, C, E, G, and I) and all-cause mortality (Panels B, D, F, H, and J), stratified by prespecified clinical categories for the PNI (A and B), GNRI (C and D), and CONUT (E and F) and by cohort-specific tertiles for the SII (G and H) and WWI (I and J). Shaded bands are 95% confidence intervals; tick marks denote censored observations. Numbers at risk (unweighted) are tabulated below each panel at 0, 2.5, 5.0, 7.5, and 10.0 years. Curves are unadjusted. CONUT, Controlling Nutritional Status; GNRI, Geriatric Nutritional Risk Index; PNI, Prognostic Nutritional Index; SII, Systemic Immune-Inflammation Index; WWI, Weight-Adjusted Waist Index. Restricted cubic-spline dose–response curves for the association of each index with cardiovascular mortality: (A) PNI, (B) GNRI, (C) CONUT, (D) log(SII), and (E) WWI. Solid lines are adjusted hazard ratios with 95% confidence bands; the reference value is the cohort median for each index. The PNI and GNRI show L-shaped associations; the CONUT shows a monotonically increasing association; log-SII shows a J-shaped association; and the WWI a near-flat curve. Knots were placed at the 5th, 35th, 65th, and 95th percentiles for the PNI, GNRI, log-transformed SII, and WWI; the CONUT used 3 knots at the 10th, 50th, and 90th percentiles, owing to its discrete distribution. Models were fully adjusted (Model 3, as defined in the footnote to Table 2). CONUT, Controlling Nutritional Status; GNRI, Geriatric Nutritional Risk Index; PNI, Prognostic Nutritional Index; SII, Systemic Immune-Inflammation Index; WWI, Weight-Adjusted Waist Index; aHR, adjusted hazard ratio; CI, confidence interval. All-cause Mortality and Comparative Model Performance All five indices were independently associated with all-cause mortality in fully adjusted models; direction and magnitude paralleled the cardiovascular-mortality results for the three nutritional indices and the SII (Table 3). WWI was a significant predictor of all-cause mortality despite its null association with cardiovascular mortality (Table 3). In comparisons of fully adjusted models (Table 4, Supplementary Fig. S1), the three nutritional indices had the highest discrimination for cardiovascular mortality (Harrell’s C-statistic, approximately 0.77), followed by SII and WWI; PNI had the lowest AIC and BIC. For all-cause mortality, GNRI had the highest C-statistic, followed by PNI and CONUT (Table 4). Confidence intervals around the C-statistics overlapped across all five indices; AIC/BIC ranking favored the nutritional indices over SII and WWI for both outcomes. Table 3.Associations of nutritional, immune-inflammatory, and anthropometric indices with all-cause mortalityIndex and comparisonModel 1Model 2Model 3aHR (95% CI)p-valueaHR (95% CI)p-valueaHR (95% CI)p-valueNutritional indices PNI, per 1-unit increase0.92 (0.90–0.95)<0.0010.93 (0.91–0.96)<0.0010.94 (0.91–0.96)<0.001 PNI, normal (>50, reference)1.00 (ref)1.00 (ref)1.00 (ref) PNI, mild risk (45–50)1.51 (1.24–1.84)<0.0011.41 (1.14–1.75)0.0011.42 (1.14–1.78)0.002 PNI, moderate-to-severe risk (<45)2.82 (1.96–4.05)<0.0012.77 (2.00–3.85)<0.0012.62 (1.93–3.56)<0.001 GNRI, per 1-unit increase0.92 (0.90–0.94)<0.0010.93 (0.90–0.95)<0.0010.93 (0.91–0.95)<0.001 GNRI, no risk (≥98, reference)1.00 (ref)1.00 (ref)1.00 (ref) GNRI, at risk (<98)2.36 (1.76–3.15)<0.0012.15 (1.60–2.90)<0.0012.14 (1.59–2.87)<0.001 CONUT, per 1-point increase1.26 (1.15–1.38)<0.0011.21 (1.11–1.32)<0.0011.21 (1.11–1.32)<0.001 CONUT, normal (0–1, reference)1.00 (ref)1.00 (ref)1.00 (ref) CONUT, abnormal (≥2)1.44 (1.16–1.79)<0.0011.29 (1.04–1.60)0.0211.30 (1.06–1.60)0.012Immune-inflammatory index log(SII), per 1-unit increase1.49 (1.23–1.80)<0.0011.41 (1.18–1.68)<0.0011.34 (1.12–1.59)0.001 SII, low tertile (reference)1.00 (ref)1.00 (ref)1.00 (ref) SII, middle tertile1.18 (0.89–1.56)0.2581.16 (0.88–1.53)0.2991.17 (0.89–1.56)0.262 SII, high tertile1.68 (1.28–2.21)<0.0011.59 (1.22–2.08)<0.0011.47 (1.14–1.91)0.003Anthropometric index WWI, per 1-unit increase1.51 (1.29–1.76)<0.0011.35 (1.16–1.56)<0.0011.25 (1.07–1.45)0.005 WWI, low tertile (reference)1.00 (ref)1.00 (ref)1.00 (ref) WWI, middle tertile1.46 (1.08–1.98)0.0151.32 (0.96–1.81)0.0871.26 (0.92–1.71)0.151 WWI, high tertile1.93 (1.48–2.50)<0.0011.63 (1.27–2.09)<0.0011.42 (1.12–1.80)0.004PNI, Prognostic Nutritional Index; GNRI, Geriatric Nutritional Risk Index; CONUT, Controlling Nutritional Status; SII, Systemic Immune-Inflammation Index; WWI, Weight-Adjusted Waist Index; aHR, adjusted hazard ratio; CI, confidence interval.Hazard ratios were estimated with survey-weighted Cox proportional hazards models, pooled across 20 multiply imputed datasets using Rubin's rules. Model 1 was unadjusted; Model 2 was adjusted for age category, sex, race/ethnicity, education, and poverty–income ratio; Model 3 was Model 2 plus smoking status, current alcohol use, hypertension, diabetes mellitus, and chronic kidney disease. The SII was natural-log-transformed because its distribution was right-skewed. Cohort-specific tertiles were used for the SII and WWI. Associations of nutritional, immune-inflammatory, and anthropometric indices with all-cause mortality PNI, Prognostic Nutritional Index; GNRI, Geriatric Nutritional Risk Index; CONUT, Controlling Nutritional Status; SII, Systemic Immune-Inflammation Index; WWI, Weight-Adjusted Waist Index; aHR, adjusted hazard ratio; CI, confidence interval. Hazard ratios were estimated with survey-weighted Cox proportional hazards models, pooled across 20 multiply imputed datasets using Rubin's rules. Model 1 was unadjusted; Model 2 was adjusted for age category, sex, race/ethnicity, education, and poverty–income ratio; Model 3 was Model 2 plus smoking status, current alcohol use, hypertension, diabetes mellitus, and chronic kidney disease. The SII was natural-log-transformed because its distribution was right-skewed. Cohort-specific tertiles were used for the SII and WWI. Table 4.Discrimination and model fit of the five fully adjusted index modelsIndexCardiovascular mortalityAll-cause mortalityC-statistic (95% CI)AICBICC-statistic (95% CI)AICBICCONUT (nutritional)0.772 (0.741–0.804)2,239.02,326.90.713 (0.689–0.736)5,680.95,768.8PNI (nutritional)0.771 (0.740–0.803)2,235.32,323.30.716 (0.693–0.740)5,665.65,753.6GNRI (nutritional)0.766 (0.733–0.799)2,242.42,330.40.722 (0.699–0.745)5,655.65,743.6SII (immune-inflammatory)0.761 (0.729–0.793)2,254.42,342.30.704 (0.681–0.728)5,689.05,777.0WWI (anthropometric)0.754 (0.722–0.786)2,261.92,349.80.697 (0.673–0.721)5,691.65,779.6CONUT, Controlling Nutritional Status; GNRI, Geriatric Nutritional Risk Index; PNI, Prognostic Nutritional Index; SII, Systemic Immune-Inflammation Index; WWI, Weight-Adjusted Waist Index; AIC, Akaike information criterion; BIC, Bayesian information criterion; CI, confidence interval.Indices are ordered by discrimination for cardiovascular mortality, from highest to lowest C-statistic. Harrell's C-statistic was pooled across 20 multiply imputed datasets using Rubin's rules; AIC and BIC are arithmetic means across the 20 imputed model fits. All models are fully adjusted (Model 3, as defined in the footnote to Table 2). Discrimination and model fit of the five fully adjusted index models CONUT, Controlling Nutritional Status; GNRI, Geriatric Nutritional Risk Index; PNI, Prognostic Nutritional Index; SII, Systemic Immune-Inflammation Index; WWI, Weight-Adjusted Waist Index; AIC, Akaike information criterion; BIC, Bayesian information criterion; CI, confidence interval. Indices are ordered by discrimination for cardiovascular mortality, from highest to lowest C-statistic. Harrell's C-statistic was pooled across 20 multiply imputed datasets using Rubin's rules; AIC and BIC are arithmetic means across the 20 imputed model fits. All models are fully adjusted (Model 3, as defined in the footnote to Table 2). Covariate Associations Among the adjusted covariates, advanced age and CKD were the strongest independent predictors of cardiovascular mortality across all five index models, with consistent direction and magnitude (Supplementary Fig. S2). Current alcohol use was consistently associated with lower hazard, and current smoking was associated with elevated hazard in the PNI and CONUT models. Sex, race/ethnicity, education, and poverty–income ratio showed no consistent independent association with cardiovascular mortality. Sensitivity Analyses Lag analyses sequentially excluding early deaths (deaths within 6, 12, and 24 months of the MEC examination) showed a temporal pattern that differed across index categories (Supplementary Tables S1, S2). The three nutritional indices retained statistical significance at every landmark for both outcomes. The SII association with cardiovascular mortality attenuated and lost significance at the 24-month landmark; WWI remained null for cardiovascular mortality at every landmark. The complete-case sensitivity analysis, restricted to the 1,228 participants with non-missing values on every multiply imputed variable, corroborated the multiple-imputation findings; direction and magnitude were preserved across all five indices for both outcomes (Supplementary Table S3). Additional adjustment of Model 3 for total energy and protein intake from the NHANES Day-1 dietary recall (n=1,698; 10-year Day-1 dietary weight) yielded estimates unchanged from the primary analysis for all five indices (Supplementary Table S4). DISCUSSION Principal Findings In this nationally representative cohort of older adults with established cardiovascular disease, PNI and GNRI showed L-shaped associations with cardiovascular mortality, whereas CONUT showed a monotonic positive association. The SII association attenuated beyond 24 months; the WWI predicted all-cause but not cardiovascular mortality. PNI gave the best model fit for cardiovascular mortality and the largest adjusted hazard (>4.5-fold) for moderate-to-severe malnutrition; CONUT had the highest discrimination. These findings are hypothesis-generating and require external validation. Comparative Performance of Nutritional Indices The identification of the PNI, GNRI, and CONUT as predictors of cardiovascular mortality is consistent with prior evidence in older and cardiovascular populations. In a NHANES chronic-kidney-disease cohort, a low PNI conferred a 67% higher all-cause and a 103% higher cardiovascular mortality risk in the lowest versus highest quartile.17) The PNI also outperformed three other nutritional indices for all-cause mortality in older patients with atrial fibrillation (AUC, 0.696)30); Chang et al.7) reported a pooled hazard ratio of 2.97 for long-term all-cause mortality with low PNI after percutaneous coronary intervention, and the PNI was the strongest of four indices in older patients with HF with preserved ejection fraction.31) CONUT discrimination paralleled Sun et al.32) in hypertensive older adults (AUC 0.778) and the meta-analysis of Yuan et al.,33) in which an elevated CONUT carried a 57% higher all-cause mortality risk in HF. GNRI findings are concordant with Dong et al.34) (59% higher all-cause mortality in HF) and Miura et al.,35) in which the GNRI outperformed functional measures for 1-year death or re-admission in older hospitalized heart-failure patients. Yoon et al.10) extended this beyond CVD, reporting that a low GNRI predicted both a steeper 5-year decline in estimated glomerular filtration rate and a more than two-fold higher all-cause mortality in older Korean adults. Together, these findings suggest that the PNI, GNRI, and CONUT capture overlapping but non-redundant facets of geriatric nutritional vulnerability across cardiovascular phenotypes. Evaluation of Immune-inflammatory and Anthropometric Indices The inflammatory SII was associated with cardiovascular mortality in our primary analysis, directionally consistent with Cheng et al.36) in hypertensive women. The association attenuated, however, after exclusion of early deaths in landmark analyses, consistent with the SII capturing acute or subacute inflammatory states (infection, decompensated HF, undiagnosed malignancy) that precede death over short horizons rather than a durable chronic-exposure burden; this matches the original derivation of the SII by Hu et al.25) as a point-in-time peripheral-blood index of systemic inflammation in hepatocellular carcinoma rather than a long-term geriatric risk marker. The PNI, GNRI, and CONUT integrate serum albumin and, where applicable, body weight or the total lymphocyte count—substrates that reflect chronic protein-energy status over months to years—whereas the SII is computed from a single complete blood count and reflects the inflammatory state on the day of phlebotomy, so the differential temporal behavior of the two index classes reflects, in part, a mismatch in the exposure window. The weight-based WWI predicted all-cause but not cardiovascular mortality, concordant with Jia et al.,37) in which each unit increase in the WWI was associated with an 82% higher odds of frailty in older adults; the WWI may therefore mark sarcopenic obesity and general biological aging rather than cardiovascular-specific pathology. The absence of a cardiovascular signal is also consistent with the obesity-paradox literature in the oldest old, in which higher BMI and waist circumference predicted lower three-year all-cause mortality38); preserved adiposity in late life may confound the predictive utility of weight-based indices for cardiovascular outcomes. The SII and WWI therefore convey information complementary to the nutritional indices, but their associations with cardiovascular mortality in older adults with CVD are narrower and, for the SII, more temporally sensitive. Pathophysiological Mechanisms and the MIA Syndrome The convergent prognostic value of the PNI, GNRI, and CONUT is consistent with the malnutrition–inflammation–atherosclerosis (MIA) framework. Each index draws on serum albumin and either the total lymphocyte count (PNI, CONUT) or body weight relative to an ideal reference (GNRI); these inputs are jointly modulated by the chronic inflammatory state of aging and advanced cardiovascular disease. Hypoalbuminemia reflects cytokine-mediated hepatic reprioritization, capillary leak, and reduced antioxidant capacity; peripheral lymphopenia is a signature of immunosenescence implicated in plaque instability and impaired infection control. The MIA construct was originally formalized in end-stage renal disease and operationalized by Mikami et al.39) using the GNRI for malnutrition, high-sensitivity C-reactive protein for inflammation, and a prior cardiovascular event for atherosclerosis, with low GNRI tracking high inflammation and severe periodontitis tracking both. Ohmure et al. extended this to cardiovascular populations, reporting that the co-presence of all three MIA factors independently predicted all-cause mortality and major adverse cardiac and cerebrovascular events after endovascular therapy for lower-extremity artery disease (hazard ratio 3.46 for death over approximately 2 years).40) The retention of nutritional-index associations after exclusion of early deaths, together with attenuation of the acute inflammatory marker, is consistent with the MIA cluster operating predominantly through a chronic, cumulative pathway. Strengths and Limitations Strengths include a nationally representative NHANES sample with National Death Index linkage and a follow-up exceeding a decade; survey-weighted analysis with multiple imputation; and restricted cubic-spline modeling of nonlinear dose–response relationships. The within-cohort comparison of three index categories provides directly comparable evidence, and sensitivity analyses (landmark, complete-case, and dietary-adjusted) converged on the same conclusions. This study has limitations. First, the cohort is heterogeneous (congestive HF, coronary heart disease, angina, myocardial infarction, stroke), and NHANES lacks cardiac-specific characterization such as left ventricular ejection fraction or natriuretic peptides; the findings should not be extrapolated to phenotype-specific decisions without validation in cohorts with detailed cardiac phenotyping. Second, the comparison between the nutritional indices and the SII is conceptually asymmetric: the former integrate a chronic exposure and the latter reflects an acute snapshot. Third, all indices were measured at a single visit, precluding evaluation of longitudinal trajectories. Finally, despite multivariable adjustment, multiple imputation, and sensitivity analyses, residual confounding cannot be excluded. Clinical Implications and Future Directions The three index categories appear complementary rather than interchangeable. The performance of the PNI, GNRI, and CONUT in landmark analyses suggests they identify a durable component of cardiovascular risk linked to chronic protein-energy and immune-nutritional dysregulation; a low PNI could prompt formal nutritional assessment, and the GNRI could complement BMI in heart-failure follow-up. The SII attenuation in 24-month lag analyses suggests that an isolated elevation should be interpreted in its acute context (infection, decompensation, active malignancy) rather than as a long-term marker. The WWI signal, restricted to all-cause mortality, may reflect general frailty rather than cardiovascular-specific risk in this population. The incremental value of these indices over standard clinical assessment is unestablished. Priority research directions are: (1) validation in phenotype-specific cohorts with cardiac biomarkers and left-ventricular imaging; (2) longitudinal evaluation of within-individual change in the PNI, GNRI, and CONUT after targeted nutritional or pharmacologic intervention; and (3) multi-domain prognostic models combining the nutritional indices with functional, frailty, and cardiac-biomarker domains. Conclusion In older adults with CVD, the PNI, GNRI, and CONUT showed nonlinear associations with cardiovascular mortality that persisted across sensitivity analyses, whereas the SII reflected an acute signal and the WWI a frailty signal. These findings do not yet support routine clinical use but justify prospective external validation of the three nutritional indices as low-cost laboratory markers of long-term cardiovascular risk in this population. Footnotes The author acknowledges the National Center for Health Statistics and Centers for Disease Control and Prevention for designing and conducting the National Health and Nutrition Examination Survey (NHANES) and for making publicly available de-identified data. I thank all NHANES participants for their valuable contributions to public health research. CONFLICT OF INTEREST The researcher claims no conflicts of interest. FUNDING None. AUTHOR CONTRIBUTIONS The author performed the design and implementation of the proposed method and read and approved the final manuscript. SUPPLEMENTARY MATERIALS Supplementary materials can be found via https://doi.org/10.4235/agmr.26.0021. Forest plot of Harrell’s C-statistic for cardiovascular mortality across the five fully adjusted index models. Points denote the pooled Harrell’s C-statistic from the fully adjusted Cox model (Model 3) for each index; horizontal lines denote the corresponding 95% confidence intervals (CIs). C-statistics were pooled across 20 multiply imputed datasets using Rubin’s rules. The fully adjusted model included age category, sex, race/ethnicity, education, poverty–income ratio (PIR), smoking status, current alcohol use, hypertension, diabetes mellitus, and chronic kidney disease (CKD). CONUT, Controlling Nutritional Status; PIR, poverty–income ratio; GNRI, Geriatric Nutritional Risk Index; SII, Systemic Immune-Inflammation Index; WWI, Weight-Adjusted Waist Index. Forest plot of independent covariate associations with cardiovascular mortality from the fully adjusted Cox model with PNI as the index. Points and horizontal lines denote adjusted hazard ratios (HRs) with 95% confidence intervals (CIs) from the fully adjusted Cox model (Model 3) with the Prognostic Nutritional Index (PNI) as the index of interest, with one row per covariate. The vertical dashed line indicates the null effect (HR=1.00). The model adjusted for the PNI and for age category, sex, race/ethnicity, education, poverty–income ratio (PIR), smoking status, current alcohol use, hypertension, diabetes mellitus, and chronic kidney disease. Landmark sensitivity analysis for cardiovascular mortality: fully adjusted hazard ratios after sequentially excluding deaths within 6, 12, and 24 months of the MEC examination Landmark sensitivity analysis for all-cause mortality: fully adjusted hazard ratios after sequentially excluding deaths within 6, 12, and 24 months of the MEC examination Complete-case sensitivity analysis (n=1,228): fully adjusted Cox models without multiple imputation for CVD and all-cause mortality Sensitivity analysis additionally adjusting Model 3 for total energy and protein intake from the NHANES Day-1 dietary recall: cardiovascular mortality REFERENCES Associated Data Supplementary Materials Forest plot of Harrell’s C-statistic for cardiovascular mortality across the five fully adjusted index models. Points denote the pooled Harrell’s C-statistic from the fully adjusted Cox model (Model 3) for each index; horizontal lines denote the corresponding 95% confidence intervals (CIs). C-statistics were pooled across 20 multiply imputed datasets using Rubin’s rules. The fully adjusted model included age category, sex, race/ethnicity, education, poverty–income ratio (PIR), smoking status, current alcohol use, hypertension, diabetes mellitus, and chronic kidney disease (CKD). CONUT, Controlling Nutritional Status; PIR, poverty–income ratio; GNRI, Geriatric Nutritional Risk Index; SII, Systemic Immune-Inflammation Index; WWI, Weight-Adjusted Waist Index. Forest plot of independent covariate associations with cardiovascular mortality from the fully adjusted Cox model with PNI as the index. Points and horizontal lines denote adjusted hazard ratios (HRs) with 95% confidence intervals (CIs) from the fully adjusted Cox model (Model 3) with the Prognostic Nutritional Index (PNI) as the index of interest, with one row per covariate. The vertical dashed line indicates the null effect (HR=1.00). The model adjusted for the PNI and for age category, sex, race/ethnicity, education, poverty–income ratio (PIR), smoking status, current alcohol use, hypertension, diabetes mellitus, and chronic kidney disease. Landmark sensitivity analysis for cardiovascular mortality: fully adjusted hazard ratios after sequentially excluding deaths within 6, 12, and 24 months of the MEC examination Landmark sensitivity analysis for all-cause mortality: fully adjusted hazard ratios after sequentially excluding deaths within 6, 12, and 24 months of the MEC examination Complete-case sensitivity analysis (n=1,228): fully adjusted Cox models without multiple imputation for CVD and all-cause mortality Sensitivity analysis additionally adjusting Model 3 for total energy and protein intake from the NHANES Day-1 dietary recall: cardiovascular mortality