Data for this study were obtained from the National Health and Nutrition Examination Survey (NHANES) [20], a program designed to generate vital and health statistics for the US population. Given that CRP data were available only for 5 consecutive survey cycles (2001‐2002, 2003‐2004, 2005‐2006, 2007‐2008, and 2009‐2010), we initially collected data from 52,195 participants across these cycles. Participants were excluded if they met any of the following criteria: (1) missing blood test results for albumin, CBC, or CRP; (2) absence of a cancer diagnosis or missing information on cancer history; (3) missing data on relevant covariates; or (4) pregnancy.

For all participants, CBC was performed on blood specimens using a Beckman Coulter MAXM instrument at the Mobile Examination Centers. CRP levels were quantified using latex-enhanced nephelometry, and albumin concentrations were measured using the DcX800 method.

The inflammatory markers were calculated using the following formulas:

Cancer diagnosis was based on self-reported medical history. Participants were identified as having cancer if they responded affirmatively to the question: "Have you ever been told by a doctor or other health professional that you had cancer or a malignancy of any kind?” Mortality status for the NHANES follow-up population was obtained from the publicly available mortality files [21]. The primary outcome was all-cause mortality, while secondary outcomes included cancer-specific and cardiovascular mortality, coded according to the International Classification of Diseases, 10th Revision (ICD-10). Follow-up time was calculated in person-months from the date of the interview, with the observation period ending on December 31, 2019.

Demographic characteristics included gender, age, race, educational level, marital status, smoking status, alcohol consumption, and the poverty-to-income ratio. Race was categorized as non-Hispanic White, non-Hispanic Black, other Hispanic, Mexican American, or other races. Educational level was classified as: less than 9th grade, 9th to 11th grade (including 12th grade with no diploma), high school graduate or General Educational Development diploma, some college or associate degree, and college graduate or above. Marital status included the following categories: married, widowed, divorced, separated, never married, and living with a partner. Participants who consumed at least 12 drinks of any type of alcoholic beverage within 1 year were defined as drinkers. Those who reported smoking at least 100 cigarettes during their lifetime were categorized as smokers. Medical history—including hypertension, diabetes, hyperlipidemia, myocardial infarction (MI), congestive heart failure (CHF), coronary heart disease (CHD), and stroke—was based on prior medical records provided by health care professionals or physicians.

The distribution of continuous variables was assessed using the Shapiro-Wilk test. Given the nonnormal distribution of the variables, they were summarized as medians with IQRs. Categorical data were expressed as frequencies and percentages. Differences among groups stratified by the 6 inflammatory markers were evaluated using the Kruskal-Wallis test or Fisher exact test, as appropriate.

To evaluate the prognostic value of the 6 inflammatory markers for mortality, weighted Cox regression analyses were performed to estimate hazard ratios (HRs) with 95% CIs. Three adjustment models were constructed: model I was unadjusted; model II was adjusted for gender, age, race, educational level, marital status, and poverty-to-income ratio; and model III was additionally adjusted for smoking, alcohol consumption, hypertension, diabetes, hyperlipidemia, CHF, CHD, MI, and stroke, in addition to the covariates included in model II. Restricted cubic spline (RCS) models were further used to explore potential dose-response relationships between the inflammatory markers and mortality outcomes, with adjustment for demographic characteristics, smoking, alcohol consumption, and disease history. Kaplan-Meier survival curves with log-rank tests were used to estimate survival distributions across tertiles of the inflammatory markers. Receiver operating characteristic curves were generated to assess the sensitivity and specificity of the 6 inflammatory markers, and the area under the curve (AUC) was calculated to evaluate their predictive performance for all-cause, cancer-specific, and cardiovascular mortality. Subgroup analyses were conducted concurrently to elucidate potential interactions between the 6 inflammatory markers and covariates. To assess the robustness of the results, a sensitivity analysis was performed excluding patients with asthma and arthritis.

All statistical analyses were performed using R software (version 4.3.0; R Core Team 2023), and a P value of less than .05 was deemed statistically significant.

The protocol for NHANES was approved by the National Center for Health Statistics Ethics Review Board (protocols: 98‐12 and 2005‐06 [22]) in accordance with the Declaration of Helsinki. All participants signed written informed consent.

A total of 946 participants were enrolled in this study (Figure S1 in Multimedia Appendix 1). Based on all-cause mortality status, participants were classified into survival and nonsurvival groups. Table 1 summarizes the weighted baseline characteristics comparing the 2 groups. The survival group had a higher proportion of women, was younger, and exhibited higher educational levels, higher income, and more favorable marital status compared with the nonsurvival group. The nonsurvival group reported a higher prevalence of hypertension, diabetes, CHF, CHD, MI, stroke, and weak or failing kidneys. Furthermore, the values of SIRI, SII, PLR, NLR, and IPI were higher in the nonsurvival group than in the survival group, whereas the CALLY index was lower in the nonsurvival group (Table 1; Figure S2 in Multimedia Appendix 1).

In Cox regression models, the levels of the 6 inflammatory markers were stratified by tertiles to evaluate their associations with all-cause, cancer-specific, and cardiovascular mortality (Table 2 and Tables S1 and S2 in Multimedia Appendix 1). A 1-SD increase in SIRI was associated with an increased risk of all-cause, cancer-specific, and cardiovascular mortality (all P<.001). Similar associations were observed for SII, NLR, and IPI, with elevations in these markers corresponding to higher mortality risks for all 3 outcomes. Associations of PLR and CALLY with all-cause and cardiovascular mortality were observed only in unadjusted or partially adjusted models, with no significant associations with cancer-specific mortality.

With regard to the primary outcome of all-cause mortality, the higher SIRI tertile was associated with an increased risk (tertile 3 vs tertile 1; hazard ratio [HR]: model I: 2.67, 95% CI 2.09‐3.41; model II: 1.74, 95% CI 1.35‐2.25; model III: 1.72, 95% CI 1.29‐2.27). The HRs for NLR and IPI in the second and third tertiles were significantly higher than those in the first (reference) tertile, with P value for trend less than .05. CALLY was the only protective factor, with HR values progressively decreasing from tertile 2 to tertile 3 compared with tertile 1 (P value for trend<.001).

Regarding cancer-specific mortality, IPI and CALLY appeared to be the most sensitive indicators. A higher IPI level was associated with an increased risk of cancer-specific mortality (tertile 3 vs tertile 1; HR: model I: 2.74, 95% CI 1.74‐4.33; model II: 2.55, 95% CI 1.59‐4.09; model III: 1.91, 95% CI 1.11‐3.27). In contrast, a higher CALLY level was associated with a decreased risk of cancer-specific mortality (tertile 3 vs tertile 1; HR, 95% CI: model I: 0.38, 0.24‐0.60; model II: 0.41, 0.25‐0.65; model III: 0.53, 0.31‐0.90). The other 4 markers were not significantly associated with cancer-specific mortality.

As shown in Table S2 in Multimedia Appendix 1, SIRI may be considered a predictor of cardiovascular mortality. Compared with tertile 1, the HR for tertile 3 was 3.50 (95% CI 1.51‐3.66) in model I, 1.86 (1.06‐3.27) in model II, and 1.83 (0.99‐3.39) in model III, with P value for trend of less than .001, .02, and .04, respectively. The other markers were not significantly associated with cardiovascular mortality.

Multivariable-adjusted RCS analyses revealed significant dose-response relationships between the 6 inflammatory markers and all-cause mortality (Figure 1). SIRI (P value for nonlinear=.51) and NLR (P value for nonlinear=.05) presented linear correlations, while SII (P value for nonlinear=.001), PLR (P value for nonlinear<.001), IPI (P value for nonlinear=.009), and CALLY (P value for nonlinear=.001) exhibited nonlinear associations.

As shown in Figure S3 in Multimedia Appendix 1, SII (P value for nonlinear<.001), PLR (P value for nonlinear<.001), NLR (P value for nonlinear<.001), IPI (P value for nonlinear<.001), and CALLY (P value for nonlinear=.008) exhibited nonlinear dose-response relationships with cancer-specific mortality, whereas SIRI did not (P value for nonlinear=.16). Regarding cardiovascular mortality, SII (P value for nonlinear=.005) and CALLY (P value for nonlinear=.02) demonstrated nonlinear dose-response relationships, whereas SIRI, NLR, and IPI did not.

Survival curves for all-cause mortality stratified by tertiles of SIRI, SII, PLR, NLR, IPI, and CALLY are presented in Figure 2A (log-rank test; P<.001). Higher levels of SIRI, SII, PLR, NLR, and IPI were associated with worse survival probability, whereas CALLY exhibited the opposite trend.

Notably, the survival trends for cancer-specific and cardiovascular mortality across the 6 inflammatory markers were similar to those observed for all-cause mortality. However, no significant difference in survival probability was observed among PLR tertile groups (Figure S4 in Multimedia Appendix 1).

Figure 2B-D presents the predictive performance of the 6 inflammatory indices for all-cause, cancer-specific, and cardiovascular mortality. For all-cause mortality, the following AUC values were observed: SIRI: 0.6526, SII: 0.5924, PLR: 0.5544, NLR: 0.6206, IPI: 0.6212, and CALLY: 0.6111. Additionally, IPI yielded the highest AUC for cancer-specific mortality, while SIRI was the most efficient predictor of cardiovascular mortality.

No significant interactions were observed between SII, PLR, or NLR and all-cause mortality (all P>.05). However, gender, age, and smoking were found to potentially interact with SIRI, IPI, and CALLY (Figure 3). Regarding cancer-specific mortality, race, educational level, and marital status exhibited prominent interactions with SIRI, PLR, and IPI (Figure S5 in Multimedia Appendix 1). Furthermore, higher SIRI, SII, and NLR levels were more strongly associated with a higher prevalence of cardiovascular mortality in the married subgroup, whereas a higher cardiovascular mortality risk associated with IPI was observed in the other group (Figure S6 in Multimedia Appendix 1).

After excluding patients with asthma, the risk of all-cause mortality showed a progressively increasing trend across tertiles of SIRI, NLR, and IPI, whereas CALLY showed a progressively decreasing trend (P value for trend<.01; Table S3 in Multimedia Appendix 1). After excluding patients with arthritis, the risk of all-cause mortality showed a progressively increasing trend across tertiles of SIRI and IPI, whereas CALLY showed a progressively decreasing trend (P value for trend<.05; Table S4 in Multimedia Appendix 1).