Ultraprocessed foods may significantly increase risk for lung cancer

The major environmental risk factors for lung cancer tend to be substances that are directly inhaled into the lungs. Of course tobacco smoke is the single dominant cause accounting for about 85% of lung cancers. Current smokers have ~20× the risk of never-smokers, with risk rising by dose (pack-years) and duration. Second-hand smoke (i.e. living or working around smokers) raises lung-cancer risk in never-smokers by ~20–30%.

Radon is a radioactive gas that accumulates indoors and can be inhaled into the lungs where the radiation can mutate lung cells. It is thought to be the number 1 cause in never-smokers and number 2 overall in the U.S., leading to an estimated ~21,000 U.S. deaths/year, although there is some debate over these numbers.  

Occupational carcinogens such as asbestos, diesel exhaust, and certain metals (e.g., chromium, nickel, beryllium, arsenic) are established causes of lung cancer that are inhaled, and risks increase with intensity and duration of exposure. 

Finally, breathing more PM2.5 (fine particulate air pollution typically arising from combustion/burning) increases lung-cancer risk. Long-term PM2.5 exposure shows ~8–16% higher lung-cancer risk per 10 µg/m³ in meta-analyses.

Thus, it was surprising that a recent study found a significant increase in lung cancer risk from the consumption of ultra-processed foods (UPFs), which are typically not inhaled into the lungs. Researchers asked whether UPF intake was associated with risk of lung cancer overall and by subtype (NSCLC, SCLC). As a reminder, Non-Small Cell Lung Cancer (NSCLC), which represents ~85-90% of all lung cancers, is somewhat less aggressive and afflicts both smokers and non-smokers. Small Cell Lung Cancer (SCLC) (~10-15% of cases) tends to be more aggressive and is typically caused by smoking.

UPFs include packaged sweets and desserts such as cookies, cakes, and ice cream; snack foods such as potato chips; soft drinks and artificially sweetened drinks; fast food, e.g. McDonald’s; and instant food such as breakfast cereals and instant noodles. In general, UPFs are composed of industrial formulations with little whole food, and possess high palatability (tasty) and extensive additives. They can contribute up to ~60% of energy intake and have been tied to multiple adverse outcomes.

The new study was a prospective analysis of a cancer screening trial (PLCO). 101,732 adults without cancer (mean age 62.5) were followed for an average of 12.2 years. The endpoint was the incidence of lung cancer. Study participants self-reported lung cancer diagnoses through annual questionnaires, and then each lung cancer case was pathologically verified and subtyped (NSCLC, SCLC).  

Every participant filled out a food questionnaire, and these food and beverage items were categorized into one of the four (NOVA, a food classification system based on industrial processing) food groups by two trained dietitians: (1) unprocessed or minimally processed foods, (2) processed culinary ingredients, (3) processed foods and (4) UPF. The data were then tabulated to estimate the UPF servings (portions) per day, which were energy-adjusted (i.e. normalized to total caloric intake) to account for differences in total food consumption between individuals.

The study population was divided into 4 quartiles based on UPF consumption with the average consumption being 2.8 (standard deviation = 2.4) energy-adjusted servings/day. Quartile 1 (Q1) was < 1.0 whereas Q4 was  ≥3.7 servings/day. Interestingly, the top UPF contributors were lunch meat (11.1%) and diet soft drinks (caffeinated 7.3%, decaffeinated 6.6%).

Importantly, the cohort explicitly included current and ever-smokers, and about 52.6% of participants were smokers. Other variables that were recorded (possible confounders contributing to lung cancer) included body mass index (BMI), physical activity, smoking status, alcohol, education, employment status, race, and family history of any cancer.

Among the 101,732 subjects, there were 1,706 lung cancers over 1.21 million person-years; NSCLC represented 86.3% and SCLC 13.7% of the cancers in line with the expected proportions. The Q4 quartile with the highest UPF consumption had 495 lung cancer cases compared to 331 for Q1, suggesting a higher risk associated with eating more UPFs.

To make the results more rigorous, a multivariable Cox regression model adjusted for other variables (i.e. confounders) produced a hazard ratio (HR) of Q4 versus Q1 (95% confidence interval in parentheses): Lung cancer, HR = 1.41 (1.22–1.60); NSCLC, HR = 1.37 (1.20–1.58); and SCLC: HR = 1.44 (1.03–2.10). The HR value of ~1.4, a roughly 40% increase in risk, was statistically significant.

Across the four quartiles, a monotonic (i.e. steadily increasing) dose-response curve was observed (see Figure 1) with more UFP servings per day associated with more lung cancers. Within the Q4 quartile (≥3.7 UPF servings/day), the curve seemed to plateau so that those who had 4 servings per day had roughly the same hazard as those with 10.

Importantly, the researchers split apart the data from smokers versus nonsmokers, and ran subgroup analyses by smoking status. They found essentially little difference between the two groups with the lung cancer hazard ratio associated with high UPF consumption (Q4) being 1.34 in current/former smokers compared to 1.44 in never smokers. Thus, there was no significant interaction by smoking status.

The authors acknowledged limitations of the study which first and foremost included residual confounding. For example, smoking intensity was not modeled (not a variable in the Cox model), and hence it is possible that high UPF consumers also smoked with greater intensity (i.e. more packs per day) than smokers in the Q1 quartile. Likewise, there could be other variables not included in the model, or variables whose interactions with the outcomes were not captured by the linear model. Finally, because the study was observational (not an RCT), one cannot conclude a causal relationship between amount of UPF eaten and lung cancer.

However, a related paper examined in a different observational study a broader spectrum of cancers with respect to UPF consumption. Hazard ratios were calculated for the incidence and mortality of 34 different cancer types. Once again a dose-response trend was observed with each 10 percentage point increase in UPF (e.g. from 2.0 to 2.2 servings/day) being associated with higher overall cancer incidence (HR = 1.02, 95% CI = 1.01 to 1.04). The trend was strongest for ovarian cancer incidence (HR = 1.19). Likewise for cancer mortality each 10% UPF increase was associated with higher overall mortality (HR = 1.06, CI = 1.03 to 1.09) with ovarian cancer once again possessing the largest signal with HR = 1.30 for each 10% increase. Using the quartile scheme, the overall cancer hazard ratios were 1.07 (incidence) and 1.17 (mortality) comparing Q4 to Q1 UPF consumption. For ovarian cancer, the difference between top and bottom quartiles were incidence HR = 1.25 (1.00 to 1.57) and mortality HR = 1.38 (1.04 to 1.82). Roughly speaking, these results were in agreement with the PLCO study described above.

In terms of mechanism, not much is known about how UPFs may contribute to lung cancer, and so one can only speculate especially since a causal relationship has not been proven as mentioned above. Given the broad effect on numerous cancers, one may suspect that consuming UPFs may influence the initiation and progression of cancer in a non-specific indirect fashion. In other words, there does not have to be direct contact with the lungs via inhalation as with most lung carcinogens.

In summary, higher UPF intake is associated with increased hazards of lung cancer overall and by subtype. These finding support the view that UPF is a potential environmental risk factor, and are consistent with other reports in the literature, such as evidence linking UPF/Western dietary patterns to cancer whereas diets high in fruits/vegetables, fish, and whole grains show a protective effect against cancer. Eventually one would like to run a randomized controlled trial (RCT) in which the "treatment" group would consume no UPFs, whereas the control group would consume their typical amount. Then the incidence of lung (or any other) cancer would be tabulated. Until then, the research community should conduct replication observational studies in diverse populations, as well as undertaking mechanistic studies to clarify pathways.

Figure 1. Dose-response between UPF consumption (x-axis) and lung cancer incidence hazard ratio (y-axis). The hazard ratio was computed with respect to a reference level of 0 servings per day (the hazard ratios in the text were from comparing Q4 to Q1). The dotted lines signify the 95% confidence index (reproduced from Fig. 3 of Wang et al. BMJ Thorax, 2025).