This prospective cohort study was based on a nationally representative sample from the United States National Center for Health Statistics National Health and Nutrition Examination Survey (NHANES). This survey utilizes stratified, multistage probability cluster sampling, and has been performed in 2-year cycles since 1999 to manage the health and nutritional status of non-institutionalized populations in the United States [21, 22]. This survey has been performed in 2-year cycles since 1999 to manage the health and nutritional status of the United States population. Informed consent was obtained from all participants, and all NHANES protocols were approved by the Ethics Review Board of the National Center for Health Statistics. This modeling investigation was exempt from review because it used published, de-identified datasets with no personal information. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement declaration.

Each participant was interviewed and asked to complete a series of physical examinations and laboratory tests at a mobile examination center. This study analyzed data on sociodemographic traits, lifestyle factors, laboratory results, and medical history in adults aged 40 or older with available CRP, albumin, and PA data from 6 NHANES cycles from 1999 to 2010.

Information on the cancer diagnosis and type was obtained through face-to-face interviews. Interviewees were asked, “Have you ever been told by a doctor or other health professional that you have cancer or a malignant tumor of any kind?” Individuals who replied yes were defined as cancer survivors and were asked, “What type of cancer was it?” Cancer types were further grouped into obesity-related and non-obesity-related cancers, as research on PA has mainly focused on obesity-related cancers [19, 20]. Cancers of the breast, colon, rectum, esophagus, gallbladder, kidney, liver, ovary, pancreas, stomach, and uterus were classified as obesity-related cancers, whereas other cancers were classified as non-obesity-related cancers [23, 24].

The CAR was calculated as CRP divided by the albumin level. Tertiles of CAR were categorized as low (< 0.034), medium (0.034–0.102), and high (≥ 0.102) because of the lack of established CAR cut-off values [25, 26].

Data were collected using the Global Physical Activity Questionnaire created by the World Health Organization (WHO) to assess different domains of PA, such as leisure-time PA, occupation, and transportation PA [27]. PA was converted to metabolic equivalent (MET) minutes of moderate-to-vigorous PA per week in accordance with the WHO analysis guidelines. The MET values for each type of exercise were obtained from the NHANES, and PA was calculated using the following formula: PA (MET-min/week) = MET × weekly frequency × duration of each PA [28]. According to the 2018 Physical Activity Guidelines for Americans, adults should engage in moderate-intensity PA for 150 min per week (equivalent to 600 MET min/wk) or vigorous-intensity activity for 75 min per week [15]. Based on the MET calculation formula and the PA requirements in the guidelines, we made the following classification of PA. Participants were classified as inactive (PA = 0), insufficiently active (0 < PA < 600 MET-min/wk), and active (PA ≥ 600 MET-min/wk), respectively.

The National Center for Health Statistics provided mortality data linked to the National Death Index through December 31, 2019, using the International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) to record the underlying cause of death. Cancer mortality was classified as death caused by malignant neoplasms (ICD-10, codes C00–C97). The duration of follow-up was defined as the interval in months from the interview date to the date of death or December 31, 2019, for participants who did not experience an event.

Based on previous research and expert clinical opinions, potential variables that could confound or modify the results were identified [19]. These variables include age, sex (man or woman), race or ethnicity (non-Hispanic White, non-Hispanic Black, Mexican American, other races [including multiracial and other Hispanic]), educational attainment (less than high school, high school or equivalent, and above high school), marital status (married, living with a partner, widowed, divorced, separated, or never married), poverty-to-income ratio (total family income divided by the poverty threshold; < 1.3, 1.3 to ≤ 3.5, ≥ 3.5) [29], body mass index (BMI; calculated based on the ratio of weight in kilograms to height in meters squared) categorized into three groups (< 25, 25.0–29.9, ≥ 30), smoking status (current, former, or never), alcohol use (never, former, or current), and the Healthy Eating Index-2015 [30] (HEI-2015; derived from the participant’s 24-h dietary recall interview). The HEI-2015 measures the quality of an individual’s overall diet, with scores ranging from 0 to 100 (worst to best). Hypertension was either self-reported by participants who had received a diagnosis from a health professional or determined by the NHANES-measured blood pressure (≥ 130 mm Hg [systolic] or ≥ 80 mm Hg [diastolic]). A history of cardiovascular disease and diabetes was self-reported by participants who had received either or both of these diagnoses from a health professional or were determined by a prescription history for medications used to treat these conditions.

This study involved a secondary analysis of publicly available datasets, and all analyses were conducted in accordance with the NHANES analytic guidelines. Continuous variables are presented as mean ± standard deviation (SD) or median (interquartile range), while categorical variables are presented as frequencies or percentages. One-way analysis of variance was used to test statistical differences among the CAR groups for continuous variables and the chi-square test for categorical variables in the baseline characteristics analysis. Additionally, some variables (poverty-to-income ratio, BMI, and HEI-2015) had missing data, and the K-nearest neighbors' method was used to interpolate the missing values. Statistical tests were two-sided, and statistical significance was set at P < 0.05.

Multivariable Cox proportional hazards regression models were used to estimate hazard ratios (HR) and 95% confidence intervals (CI) for the associations of CAR and PA with overall and cancer-specific mortality, respectively. The final-stage multivariable models were adjusted for age, sex, race, ethnicity, educational attainment, family poverty income ratio, BMI, smoking status, alcohol use, HEI-2015 score, and health factors (hypertension, history of diabetes, and cardiovascular disease). To examine joint associations, participants were classified according to their CAR and PA levels (Low CAR + Inactive PA; Low CAR + Insufficiently active PA; Low CAR + Active PA; Medium CAR + Inactive PA; Medium CAR + Insufficiently active PA; Medium CAR + Active PA; High CAR + Inactive PA; High CAR + Insufficiently active PA; High CAR + Active PA), and mortality risks were estimated using multivariable Cox proportional hazards regression models adjusted for the same set of covariates. All analyses were conducted for the overall population, as well as for women, men, and survivors of obesity- and non-obesity-related cancers. Sensitivity analyses were conducted by excluding deaths during the first 2-year follow-up to reduce the probability of reverse causation [31]. Finally, the restricted cubic splines with Cox proportional hazard models were used to depict the linearity/nonlinearity associations between InCAR and HRs of all-cause mortality.

R software (version 4.3.2; R Foundation for Statistical Computing; http://www.Rproject.org) and Free Statistics software (version 1.9; Beijing Free Clinical Medical Technology Co., Ltd.) were used for the analysis. Data analysis was performed from October 1 to December 30, 2023.