1 INTRODUCTION

Lung cancer is one of the most prevalent types of cancer worldwide, accounting for approximately 11.6% of all cancer cases and is the leading cause of cancer-related deaths, with over 1.7 million fatalities reported globally in 2018.¹ In the United States, lung cancer is estimated to cause 240,000 new cases and 130,000 deaths in 2023, and is the primary cause of cancer-related mortality in people aged 50 years or older.² Although several risk factors, including smoking, exposure to secondhand smoke, radon gas, asbestos, air pollution, and genetic predisposition, have been identified for lung cancer,^{3, 4} the incidence, and mortality rates of lung cancer have not been well controlled.

Recently, there has been increasing focus on the effect of dietary quality in reducing the risk of both cancer incidence and mortality.^5-7 A prospective study of over 250,000 individuals systematically examined the correlation between lung cancer risk and dietary quality, revealing that five distinct indices of dietary quality were related to reduced lung cancer incidence.⁷ In addition, a study based on the Health Professionals Follow-up Study (HPFS) and the Nurses' Health Study (NHS) demonstrated that four different healthy diet quality scores were correlated with a reduced risk of lung cancer mortality.⁸ However, traditional healthy dietary patterns have overlooked the importance of sustainability, which is a crucial factor in achieving the sustainable development goals outlined by the United Nations and ensuring environmental preservation within reasonable limits.⁹

In 2019, the EAT-Lancet Commission on Healthy Eating from Sustainable Food Systems developed the ELD as a specific sustainable dietary pattern recommended for promoting environmental sustainability and reducing chronic disease incidence and mortality associated with diet.⁹ This dietary pattern mainly prioritizes incorporating fruits, legumes, vegetables, nuts, and whole grains into the diet while limiting saturated fat, added sugar, and animal-source foods.⁹ In comparison with the preceding healthy plant-based diet, the ELD is a scientifically substantiated dietary pattern that considers various interrelations among health, nutrition, and the environment to estimate the health and environmental impacts of adopting an alternative diet. Importantly, the ELD provides comprehensive guidelines for sustaining the global population within planetary resource limitations and integrating sustainability criteria into national dietary recommendations for culturally diverse countries.

Previous studies have primarily focused on investigating the association between adherence to the ELD and chronic diseases such as diabetes and cardiovascular disease, with limited attention paid to cancer incidence and mortality.^10-13 To our knowledge, only one study has analyzed the association between ELD and lung cancer incidence based on the NutriNet-Santé cohort, which mainly explored the relationship of following ELD with the risk of cardiovascular disease and overall cancer incidence, with a limited sample size for lung cancer patients.¹⁴ Additionally, no studies have explored the association between ELD and the mortality of lung cancer. To fill this gap, we comprehensively investigated the potential association of adherence to ELD with the incidence and mortality of lung cancer and its subtypes in a large US population based on the data obtained from the PLCO Screening Trial.

2 METHODS

2.1 Study design

The PLCO Cancer Screening Trial is a comprehensive and multicenter prospective study that seeks to ascertain if screening tests for ovarian, prostate, colorectal, and lung cancers can effectively lower the associated death rates of these malignancies.¹⁵ From November 1993 to September 2001, the PLCO trial successfully recruited a cohort of 154,887 adult individuals between the ages of 55 and 74 years from 10 strategically located screening centers across the USA.¹⁵ The participants were randomly assigned to either a screening group or a control group. The screening group received regular screening tests for the four types of cancer, while the control group received usual medical care. Participants were followed up until 2009 for cancer incidence and in 2018 for cancer-related mortality to evaluate the potential preventive and protective impact of screening tests on targeted cancers. The PLCO trial primarily collected demographic data through a BQ and dietary-related data through a DHQ. The PLCO Screening Trial project was granted approval by the Institutional Review Board of the National Cancer Institute (NCI), as well as each screening center involved in the study, with explicit, informed, and written consent obtained from all participants. Additionally, the NCI provided approval for our present study (Project ID: PLCO-1177).

In the present study, our aim was to investigate whether adherence to ELD was associated with the incidence and mortality of lung cancer. Accordingly, 154,887 participants enrolled in the PLCO trial were further excluded based on the following criteria:

1. Failure to complete the BQ (n = 4918).
2. Failure to return DHQ responses (n = 33,241).
3. Invalid DHQ responses, defined as those lacking a completion date, those completed after the death date, those with a high frequency of missing responses (≥8), or those with extremely high energy intake values (the first or last percentile) (n = 5221).
4. Personal history of any type of cancer before the DHQ entry (n = 9684).
5. Occurrence of outcome events during the period between DHQ entry and DHQ completion, which for this study, included the development of lung cancer, death, or lack of a follow-up analysis (n = 68).

Ultimately, in total, 101,755 participants, including 49,496 men and 52,259 women, were eligible for inclusion, as depicted in Figure 1.

2.4 Outcome ascertainment

The PLCO Cancer Screening Trial primarily relied on yearly study update forms to identify cases of lung cancer. These forms were sent to surviving participants and provided questions regarding any cancer diagnoses, including the cancer type, date, and place of diagnosis, as well as the contact information of the study participant's physician. Reported cases of cancer were confirmed through medical records. Participant vital status was also verified via the yearly study update forms, with repeated attempts made to contact participants who failed to return the forms via phone or email. To ensure complete death information, the trial also routinely checked the national death index and used the International Classification of Diseases, Ninth Revision (ICD-9) to determine causes of death based on death certificates.

2.4.1 Statistical analysis

In cases in which variables had less than 5% missing data, missing values were imputed using mode and median imputation for classification and continuous variables, respectively.¹⁸ Multiple imputations methods were used to impute the continuous variable “physical activity level,” which had a missing rate of approximately 25%.¹⁹ Detailed imputation information on the types and proportions of missing data can be found in Table S2.

A Cox proportional hazards regression was applied to determine HR and 95% CI for the relationship between ELD score and the incidence and death rate of lung cancer, with the follow-up duration as the time metric. Here, for primary outcomes, namely lung cancer incidence, follow-up length was measured from the DHQ completion date to the date of lung cancer diagnosis, death, loss, or end of follow-up (the end date for the study was December 31, 2009), whatever happened first. In addition, for secondary outcomes, namely the mortality of lung cancer, follow-up duration was computed from the DHQ completion date to the date of death, loss to follow-up, or end of the study (2018), whichever occurred earlier (Figure 2). In order to test whether there was a trend between lung cancer incidence or mortality and the quartiles of ELD scores, each participant was designated a median quartile score within that quartile and subsequently treated like a continuous variable in the regression model, using the lowest quartile as the reference group. Additionally, the ELD score was directly considered a continuous variable for computing risk estimates per 1 score increment. A screening process was carried out to identify potential confounding variables through a comprehensive review of the relevant literature and clinical expertise of the researchers.^{3, 4, 20, 21} The selected covariates were subsequently incorporated into the Cox regression model to mitigate their potential influence on the outcomes. In detail, Model 1 was adjusted for demographic features, such as sex, age, education level, race, and occupational status. In Model 2, further adjustments were made for the trial arm, lifestyle factors (including smoking status, daily cigarette consumption, drinking status, alcohol consumption, energy intake from diet, BMI at baseline, BMI at age 20, and physical activity level), and health status (aspirin use history, history of diabetes, hypertension, emphysema, chronic bronchitis, and a family history of lung cancer). To depict lung cancer incidence and mortality trends across the entire range of ELD scores, a restricted cubic spline model with three knots at the 10th, 50th, and 90th percentiles was employed.²² The median of the first quartile of ELD scores (i.e., 9) was set as the reference value. The p_nonlinearity was also determined by testing the null hypothesis, which implied that the regression coefficient of the second spline was zero. Furthermore, the above methodology was employed to assess the association between ELD score and the incidence and mortality rates of lung cancer subtypes, including NSCLC and SCLC. In addition to the primary analyses, we conducted exploratory analyses to assess the association between ELD scores and incidence and mortality of the two major lung cancer histological subtypes, adenocarcinoma and squamous cell carcinoma, with the same statistical methods.

Pre-specified subgroup analyses were carried out to explore the connection between ELD score and the incidence and mortality of lung cancer in terms of age (>65 years vs. ≤65 years), gender (male vs. female), BMI at baseline (≤30 vs.>30 kg/m²), family history of lung cancer (no vs. yes/possible), smoking status (never vs. current/former), history of emphysema (no vs. yes), aspirin use (no vs. yes), and trail arm (intervention vs. control). To prevent false subgroup effects, p-values for the interaction were estimated by comparison of models with and without multiplicative interaction terms prior to carrying out the above subgroup analyses. In addition, p-values for the trends between quartiles of ELD scores and lung cancer incidence or mortality were calculated separately for each individual subgroup using the previously described methods.

Various sensitivity analyses were applied to ensure the robustness of obtained results: (1) individuals with extreme energy intake (>4000 or <500 kcal/day) or extreme BMI (top 1% and bottom 1%) were excluded; (2) excluding outcome events of interest occurring within the first 2 or 4 years of follow-up to investigate the potential impacts of reverse causation; (3) individuals with diabetes or respiratory comorbidities (including emphysema and chronic bronchitis) were excluded due to their higher risk of developing lung cancer^23-25; (4) adjustments were made to the number of cigarettes smoked for pack-years (continuous) instead of daily cigarette consumption (0, 1–20, or >20) to enhance the statistical power of the analysis; (5) repeated analysis in unimputed data cohort to ascertain whether obtained results were influenced by missing data imputation. All statistical analyses were carried out with the help of R software version 4.2.2, with a two-tailed test and p < 0.05 as a level of statistical significance.

3 RESULTS

3.2 Association between lung cancer incidence and ELD score

In total, 1706 newly diagnosed lung cancer cases, comprising 1464 NSCLC and 242 SCLC cases, were reported during a mean follow-up of 8.82 ± 1.95 years (897,809 person-years), with the overall incidence rate close to 0.19 cases per 100 person-years. Table 2 sums up the findings of multivariable and univariable Cox regression analyses on lung cancer and ELD score incidence. In comparison with participants in the lowest ELD score quartile, those in the highest quartile had a significantly reduced incidence of lung cancer post-adjustment for possible confounding variables (multivariable model: HR_{Quartile 4 vs. Quartile 1}: 0.73; 95% CI: 0.60, 0.89; p = 0.001 for trend). When the ELD score was considered a continuous variable, lung cancer incidence was also found to be inversely associated (multivariable model: HR for 1-point ELD score increment: 0.93; 95% CI: 0.90, 0.97).

TABLE 2. Hazard ratios of the association between Eat-Lancet diet score and lung cancer incidence.

Quartiles of ELD score	Cases	Person-years	Incidence rate per 100 person-years (95% confidence interval)	Hazard ratio (95% confidence interval) by Eat-Lancet diet score
				Unadjusted	Model 1a	Model 2b
Lung cancer
Quartile 1	782	355,526	0.22 (0.21, 0.24)	1.000 (reference)	1.000 (reference)	1.000 (reference)
Quartile 2	481	264,292	0.18 (0.17, 0.20)	0.83 (0.74, 0.93)	0.86 (0.76, 0.96)	0.91 (0.81, 1.02)
Quartile 3	320	188,429	0.17 (0.15, 0.19)	0.77 (0.68, 0.88)	0.83 (0.72, 0.94)	0.87 (0.76, 0.99)
Quartile 4	123	89,562	0.14 (0.12, 0.16)	0.62 (0.52, 0.75)	0.70 (0.58, 0.85)	0.73 (0.60, 0.89)
p for trend				<0.001	<0.001	0.001
1-point increment in ELD score	1706	897,809	0.19 (0.18, 0.20)	0.89 (0.86, 0.92)	0.91 (0.88, 0.95)	0.93 (0.90, 0.97)
Non–small-cell lung cancer
Quartile 1	672	355,526	0.19 (0.18, 0.20)	1.000 (reference)	1.000 (reference)	1.000 (reference)
Quartile 2	407	264,292	0.15 (0.14, 0.17)	0.81 (0.72, 0.92)	0.84 (0.74, 0.95)	0.89 (0.79, 1.01)
Quartile 3	275	188,429	0.15 (0.13, 0.16)	0.77 (0.67, 0.89)	0.82 (0.71, 0.95)	0.87 (0.75, 1.00)
Quartile 4	110	89,562	0.12 (0.10, 0.15)	0.65 (0.53, 0.79)	0.73 (0.59, 0.89)	0.75 (0.61, 0.92)
p for trend				<0.001	<0.001	0.002
1-point increment in ELD score	1464	897,809	0.16 (0.15, 0.17)	0.89 (0.86, 0.93)	0.92 (0.88, 0.95)	0.94 (0.90, 0.97)
Small-cell lung cancer
Quartile 1	110	355,526	0.03 (0.03, 0.04)	1.000 (reference)	1.000 (reference)	1.000 (reference)
Quartile 2	74	264,292	0.03 (0.02, 0.04)	0.90 (0.67, 1.21)	0.95 (0.71, 1.27)	1.00 (0.74, 1.35)
Quartile 3	45	188,429	0.02 (0.02, 0.03)	0.77 (0.54, 1.09)	0.85 (0.60, 1.20)	0.90 (0.63, 1.27)
Quartile 4	13	89,562	0.01 (0.01, 0.02)	0.47 (0.26, 0.83)	0.55 (0.31, 0.99)	0.58 (0.32, 1.04)
p for trend				0.007	0.054	0.115
1-point increment in ELD score	242	897,809	0.03 (0.02, 0.03)	0.86 (0.78, 0.95)	0.89 (0.81, 0.98)	0.91 (0.83, 1.01)

• ^a Model 1: Model 1 was controlled with age (continuous), sex (male, female), race (white, non-white), education levels (college below, college graduate, postgraduate) and occupation situation (homemaker, working, retired/unemployed).
• ^b Model 2: Model 2 was additionally controlled with BMI at baseline (continuous), BMI at age 20 (continuous), weight change (continuous), trail arm (intervention, control), smoking status (never, current or former), daily cigarette consumption (0, 1–20, >20), history of drinking alcohol (no, yes), alcohol consumption (continuous), aspirin use (no, yes), family history of lung cancer (no, yes, possibly), history of hypertension (no, yes), history of diabetes (no, yes), history of chronic bronchitis (no, yes), history of emphysema (no, yes), total energy intake (continuous) and physical activity level (continuous).

A similar inverse association was obtained when studying the relationship between ELD score and NSCLC incidence (multivariable model: HR_{Quartile 4 vs. Quartile 1}: 0.75; 95% CI: 0.61, 0.92; p = 0.002 for trend, HR for 1-point ELD score increment: 0.94; 95% CI: 0.90, 0.97). Nevertheless, the correlation between ELD score and SCLC incidence in multivariable analyses was not significant (multivariable model: HR_{Quartile 4 vs. Quartile 1}: 0.58; 95% CI: 0.32, 1.04; p = 0.115 for trend, HR for 1-point ELD score increment: 0.91; 95% CI: 0.83, 1.01). The restricted cubic spline model demonstrated a linear inverse dose–response association of ELD score with lung cancer and NSCLC incidence (all p > 0.05 for nonlinearity; Figure 3A,B).

In additional exploratory analyses, no significant associations were observed between ELD scores and the incidence of lung adenocarcinoma or squamous cell carcinoma (Table S3). However, the restricted cubic spline model revealed a negative linear trend between higher ELD scores and squamous cell carcinoma incidence, which was not seen for adenocarcinoma (Figure S1A,B).

The subgroup analyses demonstrated that the relationship between lung cancer incidence and ELD score was not influenced by age, sex, BMI at baseline, a family history of lung cancer, aspirin use, smoking status, history of emphysema, or trial arm (all p > 0.05 for interaction; Table S4). Moreover, in sensitivity analyses, the initial correlation between ELD score and lung cancer incidence remained materially unchanged (Table S5).

4 DISCUSSION

This study sought to explore the relationship between the ELD score and lung cancer incidence and death rate as per data from the PLCO Cancer Screening Trial. The study participants were divided into quartiles based on ELD scores ranging from 4 to 13 points. The results indicated that higher ELD scores were associated with lower incidence and mortality rates of lung cancer, particularly for NSCLC. However, this association was not observed for SCLC. Furthermore, the relationship between lung cancer incidence and mortality with ELD score was not affected by lifestyle or demographic factors, indicating that the association is independent. An inverse linear trend was observed in the dose–response analysis, indicating a decrease in lung cancer incidence and mortality with increasing ELD score. Sensitivity analysis confirmed the robustness of these findings. Therefore, adherence to the ELD, as measured by the ELD score, may reduce the incidence and mortality of lung cancer, particularly for NSCLC.

So far, there has been only one study that investigated the link between ELD and lung cancer risk.¹⁴ In this study, Berthy et al. mainly investigated the correlation between ELD and cardiovascular disease (CVD) and overall cancer risk using a modified ELD score developed by Kesse-Guyot et al.^{14, 26} The study utilized a cohort of more than 60,000 participants in the NutriNet-Santé study. However, ELD and the risk of overall cancer and CVD were not found to be significant, neither combined nor separately. Of note, Berthy et al.¹⁴ demonstrated an inverse association between ELD score and lung cancer risk in the secondary outcome analysis of their study, but this association was found to be statistically insignificant after adjustment for energy intake and BMI. It should be noted that the representativeness of the Berthy et al.¹⁴ study population was limited, as NutriNet-Santé participants were typically younger, more educated, healthier, and predominantly female compared with the general population of France. Thus, the observed associations may have been underestimated. In our study, 101,755 American adults from the PLCO trial were chosen as the study population. The PLCO participants were included according to gender equality, and no significant differences with the general population of the USA have been reported. Our study demonstrated a significant linear association of ELD score with lung cancer risk, which was limited to NSCLC. These findings provide novel insights into the effect of ELD on the occurrence of cancers in the lungs.

Although the correlation of overall mortality risk with ELD score has been reported in previous literature, the potential relationship between ELD score and lung cancer mortality has not been explored. Knuppel et al.¹⁷ evaluated the connection between mortality risk and ELD score in a large prospective cohort of UK adults but failed to find a clear association. Conversely, Stubbendorff et al.¹³ conducted a study in Sweden and found that maximum adherence to ELD reduced the risk of all-cause and cancer-related mortality by 25% and 24%, respectively. Additionally, Zagmutt et al.²⁷ questioned the efficacy of ELD as a recommended dietary guideline and emphasized the importance of considering BMI and energy intake when assessing the impact of dietary factors on mortality. Our present study comprehensively accounted for weight variables and energy intake in the analysis and incorporated them into the Cox regression model for adjustment. The obtained findings provide the first proof of a negative correlation between lung cancer mortality and ELD score, suggesting that adherence to ELD may be an effective means of reducing lung cancer mortality. Interestingly, the subtype analysis results demonstrated that the notable negative correlation of ELD score with lung cancer mortality is exclusive to NSCLC.

Actually, several previous different indices reflecting the recommendations for a healthy plant-based diet have partially support our research findings. For instance, in a multiethnic cohort study that involved 215,000 adults, the Healthy Eating Index-2015 (HEI-2015), Alternative Healthy Eating Index-2010 (AHEI-2010), and Alternative Mediterranean Diet (AMED) were significantly linked to a reduced risk of lung cancer.⁷ Moreover, the results from NHS and HPFS indicate that HEI-2015, AHEI, AMED, and the Healthful Plant-based Diet Index (HPDI) are all associated with a decreased risk of lung cancer-related death in both males and females.⁸ When comparing the composition of ELD with those of the aforementioned dietary patterns, it was found that vegetables, fruits, fiber, and red meat are the main overlapping components. These components may relate to mechanisms underlying the decreased lung cancer incidence and mortality seen. Accumulating evidence shows vegetables and fruits linked to reduced lung cancer risk,^28-30 probably due to abundant bioactive compounds such as flavonoids that can scavenge free radicals, repair DNA damage, and inhibit carcinogenesis.^{31, 32} One prospective PLCO study found higher intake of dietary fiber was associated with a 38% lower lung cancer risk.³³ The protective fiber effect may involve gut microbes producing short-chain fatty acids through fermentation, regulating insulin, anti-inflammatory pathways, cytokine production, and immune responses in anticancer defenses.^{34, 35} Additionally, limiting red meat intake decreases gut microbial carcinogen production and lowers animal fat-induced inflammation, inhibiting the release of tumorigenic factors, which could reduce lung cancer risk.^{36, 37}

This study has several significant strengths. Firstly, it is based on a large prospective cohort of over 150,000 individuals from diverse occupations recruited from 10 screening centers all over the USA. They were also followed up for an extended period of time, thereby increasing the generalizability of these findings to similar populations. Secondly, it comprehensively adjusted for various possible confounding factors in the analyses. More importantly, this is by far the first study to systematically and robustly explore the relationship between the incidence and mortality of lung cancer and its subtypes with ELD scores. The obtained findings provide strong evidence that adhering to the ELD is effective in decreasing both the incidence and mortality of lung cancer and these results are robust even after conducting a series of sensitivity analyses.

The study also has certain limitations that should be considered. For instance, using a single dietary assessment at baseline can potentially introduce some bias because of the potential changes in dietary habits over time. However, published literature has demonstrated that the correlation between diet and disease incidence using only baseline data is generally weaker than when cumulative dietary intake is used,³⁸ thereby decreasing the influence of this potential bias. Second, the self-reported FFQ may be subject to recall bias, thus affecting the accuracy of dietary information. Additionally, the study population consisted of older Americans with a mean age of 65.5 years, and it is unclear whether the observed association between ELD and risk of lung cancer incidence and mortality would hold in other regions or age groups. Therefore, future studies should explore this association in other populations to verify the generalizability of these findings.

In conclusion, this study utilized the ELD score developed by Knuppel et al. to investigate adherence to ELD among the United States population. Moreover, it revealed a linear dose–response association between ELD score and decreased risk of lung cancer incidence and mortality. Notably, this correlation was observed exclusively in NSCLC upon performing subtype analysis. Hence, these findings present a novel dietary approach to reducing both lung cancer incidence and mortality and establish an evidence-based foundation for the development of sustainable dietary guidelines and policies.

AUTHOR CONTRIBUTIONS

Yi Xiao, Ling Xiang, Yaxu Wang, and Linglong Peng contributed to the study design and data analysis. Yi Xiao and Linglong Peng contributed to the data interpretation and writing of the manuscript. Yi Xiao, Ling Xiang, Zhiquan Xu, Yunhao Tang, Hongmei He and Haitao Gu contributed to the data collection, and data curation of the present analysis. Linglong Peng, Yaxu Wang, and Haitao Gu assisted with statistical analysis and funding acquisition. All of the authors reviewed or revised the manuscript. All authors contributed to the article and approved the submitted manuscript. The work reported in the paper has been performed by the authors, unless clearly specified in the text.

FUNDING INFORMATION

This work was supported by The General Project of Chongqing Natural Science Foundation, Chongqing Science and Technology Commission, China [cstc2021jcyj-msxmX0153 (Linglong Peng)], [cstc2021jcyj-msxmX0112 (Yaxu Wang)], and [CSTB2022NSCQ-MSX1005 (Haitao Gu)], and Kuanren Talents Project of the Second Affiliated Hospital of Chongqing Medical University in China [kryc-yq-2110 (Haitao Gu)]. The funders had no role in the study design or implementation; data collection, management, analysis, or interpretation; manuscript preparation, review or approval; or the decision to submit the manuscript for publication.

CONFLICT OF INTEREST STATEMENT

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

ETHICS STATEMENT

Approval of the research protocol by an Institutional Reviewer Board: N/A.

Informed Consent: N/A.

Registry and the Registration No. of the study/trial: N/A.

Animal Studies: N/A.