Data on patients with CHD were extracted from the National Health and Nutrition Examination Surveys (NHANES) database from 2003 to 2018 in this retrospective cohort study. NHANES is conducted by the Centers for Disease Control and Prevention (CDC) and the National Center for Health Statistics (NCHS). It uses a complex, multistage stratified probability sample based on selected counties, blocks, households, and persons within households to assess the nutritional and health status of the non-institutionalized population in the United States. Interviews in participants’ homes were conducted by the NCHS-trained professionals, and extensive physical examinations including blood and urine collection were conducted at mobile exam centers (MECs). Details of the study implementation are available online: https://www.cdc.gov/nchs/nhanes/index.htm. In addition, the raw data extracted from the database are shown in the Supplementary Material.

A total of 1,908 adult patients with CHD were initially included. The exclusion criteria were missing information on (1) HALP score, (2) survival, (3) HB, (4) ALB, (5) lymphocytes, and (6) platelets. Finally, 1,633 of them were eligible. The NHANES survey was approved by the institutional review board (IRB) of NCHS. Written informed consent for participation has been obtained before every survey in NHANES. Since the data were deidentified and publicly available, no ethical approval of our IRB was required.

Diagnosis of CHD was according to the question in the NHANES multiple-choice questionnaire (MCQ160C): “Has a doctor or other health professional ever told you that you had CHD?” The participants who had positive answers were classified as patients with CHD.

In NHANES, a blood sample was obtained through examinations at the MECs, and HALP score-related indexes including serum HB, ALB, lymphocyte, and platelet levels were further analyzed in the laboratory. The HALP score was calculated according to the formula: HB (g/l) × ALB (g/l) × lymphocytes (10⁹/l)/platelets (10⁹/l) due to the fact that the unit of ALB in the database was “g/dl”. We divided the HALP score into four levels according to the quartiles (<37.31, 37.31–51.15, 51.15–69.68, and >69.68) on the basis of previous studies (18, 21).

We collected potential confounding factors including age, gender, race, education level, marital status, poverty income ratio (PIR), drinking, smoking, physical activity, total energy intake, congestive heart failure (CHF), stroke, hypertension, dyslipidemia, diabetes mellitus (DM), chronic kidney diseases (CKD), liver disease, CVD family history, cancer or malignancy, body mass index (BMI), white blood cell (WBC), monocyte, drug for CVD use, treatment for anemia, anticoagulants drug use, antiplatelet agents use, anticonvulsants use, and adrenal cortical steroids use.

The smoking status was self-reported during the NHANES household interview. Participants who claimed to have smoked fewer than 100 cigarettes in their lives were labeled ‘never smokers’. Smokers were individuals who had smoked more than 100 cigarettes in their lives (22). Drinking status was self-reported based on the question: “Had at least 12 alcohol drinks/1 year?”. Physical activity was defined by the question: “Performing high-intensity or moderate-intensity physical activity” and answered “yes” (23). Hypertension was defined by laboratory inspection, self-reported hypertension, current use of hypotensive drugs, a measured systolic blood pressure (SBP) ≥130 mm Hg or diastolic blood pressure (DBP) ≥90 mm Hg, diagnosed with hypertension, or taking antihypertensive medication. Participants with total cholesterol (TC) ≥200 mg/dl (5.2 mmol/L), triglycerides (TG) ≥150 mg/dl (1.7 mmol/L), low-density lipoprotein cholesterol (LDL-C) ≥130 mg/dl (3.4 mmol/L), HDL-C ≤40 mg/dl (1.0 mmol/L), self-reported hypercholesteremia, or receiving lipid-lowering therapy were identified as dyslipidemia. The diagnosis of DM was fasting blood glucose ≥7.0 mmol/L, glycosylated HB (HbAlc) ≥6.5%, self-reported DM, or receiving hypoglycemic therapy (24). The eGFR was calculated as follows: estimated GFR = 175 × standardized Scr −1.154 × age−0.203 × 1.212 (if Black) × 0.742 (if female), where GFR is expressed as ml/min/1.73 m² of body surface area 41 and Scr is expressed in mg/dl. The total energy intake was calculated according to the first 24-hour dietary recall in the NHANES.

The study outcome was all-cause mortality, which was identified using the international classification of diseases, ninth revisions (ICD-9) and ICD-10 codes. In NHANES, the information on death was obtained through the link of the National Death Index (NDI): https://www.cdc.gov/nchs/data-linkage/mortality.htm. The follow-up ended after participants died or up to 31 December 2019.

Measurement data were described using mean ± standard error (mean ± SE) and analysis of variance (ANOVA) for comparison. Enumeration data were expressed as numbers with constituent ratio [N (%)] and the chi-square test or Fisher's exact test for the comparison.

We used a set of NHANES special weights: the 2-year cycle of MEC exam weight (wtmec2yr) for statistical analyses because the HALP score was calculated using four laboratory indexes. Weighted univariate COX proportional hazard model analyses were used to screen the covariates. The association between the HALP score and the risk of all-cause mortality was explored using univariate and multivariate COX proportional hazard model analyses. The model was adjusted for age, education level, PIR, marital status, smoking, physical activity, total energy intake, CHF, stroke, hypertension, DM, CKD, cancer or malignancy, monocyte, drug for CVD, treatment for anemia, anticoagulants drug, and adrenal cortical steroids. We also drew a Kaplan–Meier (KM) curve to assess the survival probability of patients at different HALP levels. The receiver operator characteristic (ROC) curves were employed to reflect the predictive performance of HALP on one-year all-cause mortality in patients with CHD and compared with HB, ALB, lymphocyte, and platelet. In addition, subgroup analyses were performed to explore this relationship in patients with different ages and CHF statuses.

The evaluation indexes were hazard ratios (HRs) with 95% confidence intervals (CIs). Two-sided P < 0.05 was considered significant. The missing data are shown in Supplementary Table S1, and we employed multiple imputations to handle missing data and minimize bias. Sensitivity analysis of the association between HALP levels and all-cause mortality in CHD patients before and after the imputations of missing data are shown in Supplementary Table S2. Statistical analysis was performed using SAS 9.4 (SAS Institute, Cary, NC, USA) and R version 4.2.2 (Institute for Statistics and Mathematics, Vienna, Austria).

Figure 1 shows the flowchart of the participants screening. We initially included 1,908 adult patients with CHD from the database. Those who had missing information on HALP (n = 228), survival (n = 1), HB (n = 11), ALB (n = 11), lymphocyte (n = 8), and platelet (n = 16) were excluded. Finally, 1,633 of them were eligible.

After an average of 81 months of follow-up, 657 (33.90%) of the patients died. Their characteristics are shown in Table 1. Most of the patients were ≥65 years old (63.71%) and were men (66.05%). The average HALP score was 56.32 among participants. We compared patients with CHD who had different HALP levels and found that age, gender, smoking, dyslipidemia, CKD, HB, ALB, lymphocyte, platelet, WBC, monocyte, treatment for anemia, anticonvulsants, and adrenal cortical steroids were all significantly different among the four HALP score groups (all P < 0.05). Based on these findings, we further plotted the KM curves of the survival probability of patients with CHD at different HALP score levels (Figure 2), which showed that lower HALP scores may be associated with a lower survival probability in patients with CHD.

Supplementary Table S3 shows the covariates related to all-cause mortality, including age, education level, PIR, marital status, smoking, physical activity, total energy intake, CHF, stroke, hypertension, DM, CKD, cancer or malignancy, monocyte, drug for CVD, treatment for anemia, anticoagulants drug, and adrenal cortical steroids (all P < 0.05).

Then, we explored the relationship between HALP score and all-cause mortality in patients with CHD (Table 2). After adjusting for the covariates, we found that the HALP score was negatively associated with the risk of all-cause mortality [HR = 0.83, 95% CI: (0.74–0.93)]. Compared with patients whose HALP scores were >69.68, those who had HALP scores ≤69.68 seemed to have a high risk of all-cause mortality [HALP score of 51.15–69.68: HR = 1.36, 95% CI: (1.01–1.84); HALP score of 37.31–51.15: HR = 1.41, 95% CI: (1.07–1.85); and HALP score <37.31: HR = 1.64, 95% CI: (1.24–2.17)].

Since the HALP score was calculated by HB, ALB, lymphocyte, and platelet, we compared the predictive performance of HALP and the above four serum indexes on all-cause mortality (Figure 3). The ROCs showed that the HALP score has an AUC of 0.610 and that of HB, ALB, lymphocyte, and platelet were 0.560, 0.564, 0.595, and 0.515, respectively. The predictive values of these indexes are shown in Table 3. HALP score had a sensitivity of 0.510 and specificity of 0.654 on CHD prognosis, indicating a potential predictive value.

We further explored this association in age and CHF subgroups (Table 4). In patients who are <65 years old, we have not found an association between the HALP score and the risk of all-cause mortality (all P > 0.05). Nevertheless, those who were ≥65 years old with a HALP score of <37.31 may have an increased risk of all-cause mortality [HR = 1.86, 95% CI: (1.38–2.51)]. Furthermore, patients with [HR = 2.00, 95% CI: (1.29–3.10)] or without [HR = 1.48, 95% CI: (1.01–2.17)] CHF having HALP scores less than 37.31 both had an increased risk of all-cause mortality.