Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2022 Nov 1.
Published in final edited form as: Clin Nutr. 2021 Sep 17;40(11):5419–5429. doi: 10.1016/j.clnu.2021.09.014

Unrestrained eating behavior and risk of mortality: a prospective cohort study

Yin Zhang 1,2,3,*, Mingyang Song 3,4,5,6, Chen Yuan 1, Andrew T Chan 2,5,7,8, Eva S Schernhammer 2,6,9, Brian M Wolpin 1, Meir J Stampfer 2,3,6, Jeffrey A Meyerhardt 1, Charles S Fuchs 10,11, Susan B Roberts 12, Eric B Rimm 2,3,6, Walter C Willett 2,3,6, Frank B Hu 2,3,6, Edward L Giovannucci 2,3,6,†, Kimmie Ng 1,†
PMCID: PMC8571025  NIHMSID: NIHMS1741204  PMID: 34653818

Abstract

Background & Aims

Unrestrained eating behavior has been thought to be a proxy for diet frequency, timing, and caloric intake. We investigated the association of unrestrained eating with mortality risk in the Nurses’ Health Study prospectively.

Methods

During follow-up (1994–2016), 21953 deaths were documented among 63999 eligible participants in analyses of eating anything at any time, 22120 deaths were documented among 65839 participants in analyses of no concern with figure change. Hazard ratios (HRs) and 95% confidence intervals (CIs) were calculated using Cox proportional hazards models.

Results

Eating anything at any time was associated with an increased mortality from cancer (overall HR, 95%CI: 1.07, 1.00–1.13; driven by gastrointestinal tract cancer: 1.30, 1.10–1.54) and respiratory disease (1.16, 1.05–1.29), and decreased cardiovascular disease-specific mortality (0.92, 0.86–0.99), compared to those without this behavior; however, no association was observed between this behavior and all-cause mortality (1.02, 0.99–1.05). Women who reported having no concern with figure change experienced higher risk of mortality from all-cause (1.08, 1.05–1.11), cancer (1.08, 1.02–1.14), and respiratory disease (1.18, 1.08–1.30), compared to those not reporting this behavior. Their combined effect was associated with a higher all-cause (1.09, 1.04–1.14), cancer-specific (overall: 1.18, 1.09–1.28; gastrointestinal tract cancer: 1.36, 1.08–1.71; lung cancer: 1.09; 1.04–1.14), and respiratory disease-specific (1.30, 1.13–1.50) mortality, and was inversely associated with cardiovascular disease-specific mortality (0.88, 0.80–0.98), compared to those exhibiting the opposite.

Conclusions

Unrestrained eating was associated with increased risk of all-cause, cancer-specific (particularly for gastrointestinal tract cancer and lung cancer), and respiratory disease-specific mortality, and decreased risk of cardiovascular disease-specific mortality.

Keywords: Unrestrained eating behavior, all-cause mortality, cause-specific mortality

Introduction

Major modifiable lifestyle factors (e.g., body weight,[1–5] smoking,[1–5] alcohol intake,[1–5] diet quality,[1, 3–7] physical activity,[1–6] and sedentary behavior,[6] etc.) and their combined effects[1–5] have been associated with all-cause and cause-specific mortality[1, 2, 4, 6, 7] and life expectancy.[3, 5] In addition to diet quality, unrestrained eating behavior, as a potential proxy for diet frequency, timing (typical/inappropriate times with regard to the circadian system, and related alterations in circadian rhythms or diurnal preference), and caloric intake (with restriction or reckless abandon), has been questioned to play an independent role in these associations.

Although very few studies have explored the association between several related eating habits (eating frequency, snacking) and mortality risk, conflicting findings have been observed.[8, 9] The interpretation of their evidence may have been limited by relatively inadequate sample size and follow-up,[8, 9] incomplete assessment of eating behaviors,[8, 9] incomplete confounding control,[9] and lack of consideration for potential reverse causation.[8, 9] There has been no epidemiological study comprehensively characterizing the effect of unrestrained eating on mortality risk.

With large sample size, long-term follow-up, extensive and validated data on potential confounders,[10–15] and assessments of diverse domains of unrestrained eating exposure, the Nurses’ Health Study (NHS) longitudinal cohort[16–19] affords a rich resource to add high-quality epidemiological evidence to this topic. We prospectively investigated unrestrained eating behavior in relation to all-cause and cause-specific mortality within the NHS.

Methods

Study Population

The NHS was established in 1976 and enrolled 121,700 US female registered nurses ages 30 to 55 years, the details of which have been described previously.[16–19] Self-administered biennial questionnaires were sent to participants to update lifestyle and medical information throughout follow-up, with response rates exceeding 90% having been achieved. The baseline was set at 1994 when unrestrained eating behaviors were first assessed. Participants who were alive and free of cancer and cardiovascular disease at baseline, had no missing information on the exposures of interest, and with no implausible energy intakes (<500 or >3500 kcal per day) were eligible for inclusion.

Ascertainment of Exposures

In the 1994 NHS questionnaire, participants answered two questions on whether “I eat anything I want, anytime I want” and “I pay a great deal of attention to changes in my figure” applied to them, with “yes” or “no” as response categories. These questions, using stepwise regression, had previously been identified to be the strongest predictors of the total dietary restraint score in a study that completed a detailed questionnaire measuring three dimensions (dietary restraint, disinhibition, and hunger) of human eating behavior.[20, 21] The second question “I pay a great deal of attention to changes in my figure” is part of the “flexible, but not rigid control’ subscale proposed for dietary restraint, which is considered to have greater benefits for counterbalancing dietary disinhibition than other aspects of dietary restraint (attenuate the influence of habitual disinhibition on weight gain and body mass index (BMI)).[22] We examined each question separately and together.

Ascertainment of Deaths

Deaths were confirmed through state vital statistics records, regular searches of the National Death Index, and next-of-kin or postal authority reporting, with an identifying rate of more than 98%.[23, 24] Cohort investigators reviewed death certificates and medical records after permission from next of kin of dead participants, and classified the causes of death according to the International Classification of Diseases, Eighth Revision (ICD-8).

Ascertainment of Covariates

Throughout follow-up, participants biennially reported body weight (which we used to calculate participants’ BMI with height), smoking behavior (to calculate pack-years of smoking and time since quit smoking), menopausal status, history of aspirin, multivitamin, and postmenopausal hormone use, history of hypertension, hypercholesterolemia and diabetes, antidepressant medication use (selective serotonin reuptake inhibitors and other antidepressants) and physician-diagnosed depression. Dietary data, physical activity (metabolic equivalent of tasks (MET) scores were assigned to every specific type of physical activity, and total physical activity in MET-hours/week was calculated), and family history of cancer and cardiovascular disease were updated quadrennially. Depressive symptoms were assessed quadrennially using the 5-item Mental Health Index (MHI-5) from the Short-Form 36 Health Status Survey.[25] The Alternative Healthy Eating Index (AHEI) was calculated based on food frequency questionnaire data to measure overall dietary quality.[26] The empirical dietary index for hyperinsulinemia (EDIH) was calculated as a measurement of the potential to stimulate insulin secretion (higher scores indicating hyperinsulinemic diets).[27, 28] Diagnosis of depression was defined based on self-reported physician-diagnosed depression, the MHI-5 [25] (score≤52),[29, 30] and antidepressant medication use. Except for the above-mentioned time-varying variables, height, race, BMI at age 18, history of rotating night shiftwork, socioeconomic status (husband’s educational level), and waist circumference were captured once at the initial or subsequent questionnaires before the analytic cohort baseline. Meal frequency and breakfast consumption were assessed in 2002. The validity and reproducibility of information on anthropometrics, lifestyle, diet, and disease outcomes have been reported.[10, 12–14, 26, 31]

Statistical Analysis

Person-years of follow-up accrued from the return date of the 1994 questionnaire until the date of death recorded, loss to follow-up, or follow-up completion (defined as June 30, 2016), whichever occurred earliest.

Cox proportional hazards regression models were used to estimate hazard ratios (HRs) and 95% confidence intervals (95% CIs) associated with unrestrained eating and all-cause and cause-specific (cancer, cardiovascular disease, and respiratory disease) mortality risk. The proportionality assumption was verified using interactions between the exposures of interest and the (log-)time scale. We controlled for age (continuous, months) and follow-up cycle (each 2-year interval) in age-adjusted models. In multivariable analyses, we additionally adjusted a priori for a wide spectrum of covariates, including race (White, Black, other), cumulative average BMI (<20.0, 20.0–24.9, 25.0–29.9, and ≥30.0 kg/m2, implementing a 4-year lag-time to account for potential reverse causation), BMI at age 18 (<20.0, 20.0–24.9, 25.0–29.9, and ≥30.0 kg/m2), smoking status (never, past smoker ≥10 years since quit smoking, past smoker <10 years since quit smoking, current smoker <25 cigs/day, current smoker ≥25 cigs/day), pack-years of smoking (0, >0 to 5, >5 to 15, and >15 pack-years), alcohol intake (0, 0.1–4.9, 5–14.9, and ≥15 g/day), physical activity (MET-hours/week, continuous), total calories intake (kcal/day, quintiles), EDIH (quintiles), regular use of aspirin (yes, no), multivitamin use (yes, no), menopausal status (premenopausal, postmenopausal), postmenopausal hormone use (no use, past use, current use), history of hypertension (yes, no), history of hypercholesterolemia (yes, no, implementing a 4-year lag-time), history of diabetes (yes, no, implementing a 4-year lag-time), family history of cancer (yes, no), and family history of cardiovascular disease (yes, no).

In sensitivity analyses, 4-, 8-, and 12-year latency analyses were performed to minimize the potential influence of reverse causation on our results, and to examine whether these exposure assessments may reflect a long-term eating behavior status. We further tested various combinations of covariates in multivariable models, e.g., by removing cumulative average BMI or EDIH, replacing EDIH with AHEI (quintiles), or additionally adjusting for history of depression (yes, no), history of rotating night shiftwork (0, 1–14, and ≥15 years), socioeconomic status (husbands’ educational level: less than high school, some high school, high school graduate, college, graduate school), waist circumference (continuous, cm), meal frequency (1–2, 3–4, 5 or more times/day), and breakfast consumption (0–2, 3–5, 6–7 times/week), to comprehensively investigate how the specific adjustment for these factors affect the results (listed in table footnotes). To address the possibility of residual confounding, we estimated the predicted probability of reporting unrestrained eating using logistic regression models that included all the covariates (grouped into 3 categories: lifestyle, diet, and disease history), and additionally adjusted for the propensity score. Given the results observed for mortality from respiratory disease and lung cancer, to be prudent, our main analyses for all endpoints were then repeated specifically among participants smoked ≤ 3 lifetime pack-years to further validate the robustness of these findings.

Analyses were conducted using SAS software (version 9.4 for UNIX; SAS Institute Inc., Cary, NC), with 2-sided P values <.05 indicating statistical significance for all tests.

Results

Population Characteristics

During up to 22 years of follow-up, we documented 21953 deaths among 63999 eligible participants in analyses of eating anything at any time, 22120 deaths among 65839 eligible participants in analyses of having no concern with their figure change, and 13313 deaths among 40461 eligible participants when examining their combined effect. The sample size differs across these analyses due to the different missingness of exposures. Individuals who “ate anything at any time but cared about their figure change”, and those who “did not eat anything at any time but did not care about their figure change” were excluded specifically in analyses of the combined effect of these two unrestrained eating behaviors to have a clear contrast. Variations were observed across exposures for BMI, waist circumference, smoking status, pack-years of smoking, physical activity, overall diet quality, dietary insulinemic potential, total energy intake, meal frequency, breakfast skipping behavior, menopausal status, use of multivitamin and postmenopausal hormone, history of hypertension, hypercholesterolemia, diabetes, depression, and rotating night shift work, family history of cancer and cardiovascular disease, and husbands’ educational level. (Table 1)

Table 1.

Characteristics a of the study population in the NHS (1994) across self-reported unrestrained eating b,c

Eating anything at any time No concern with figure change Eating anything at any time + no concern with figure change
Characteristic No
n=49333		 Yes
n=14666		 No
n=42423		 Yes
n=23416		 No i
n=32890		 yes
n=7571
Age, years, mean (SD) 60.3 (7.1) 59.9 (7.2) 60.2 (7.1) 60.1 (7.2) 60.3 (7.1) 59.7 (7.3)
Race
 White 97.7 97.0 97.4 98.0 97.4 97.2
 Black 1.4 1.4 1.6 0.9 1.6 1.1
 Other 0.9 1.6 1.0 1.1 1.0 1.6
BMI, kg/m2, mean (SD) d,e 26.3 (4.9) 27.1 (6.1) 25.3 (4.3) 28 (5.9) 25.3 (4.2) 27.9 (6.3)
BMI at age 18, kg/m2, mean (SD) d 21.3 (2.9) 21.2 (3) 21.1 (2.8) 21.6 (3.1) 21.2 (2.8) 21.4 (3.1)
Waist circumference, cm, mean (SD) 86.8 (11.7) 88.4 (13.1) 85.2 (11.2) 89.9 (12.7) 85.3 (11.1) 89.9 (13.4)
Smoking status
 Never smoked 45.4 40.8 44.6 44.3 45.1 40.7
 Past smoker 43.6 36.9 44.1 40.1 44.5 36.7
 Current smoker<25 cigs/day 9.2 17.5 9.6 12.4 8.8 17.4
 Current smoker≥25 cigs/day 1.8 4.7 1.7 3.3 1.5 5.2
Pack-years of smoking, mean (SD) f 22.6 (19.8) 29.5 (23.1) 22.2 (19.5) 26.4 (22.0) 21.7 (19.2) 30.1 (23.5)
Alcohol intake, g/day, mean (SD) 5.0 (8.9) 5.2 (10.2) 5.2 (8.8) 4.8 (9.5) 5.2 (8.8) 5.2 (10.5)
Physical activity, MET-hours/week, mean (SD) g 21.2 (24.6) 17.7 (23.5) 23 (26.2) 17 (21.6) 23.1 (25.9) 16.4 (22.2)
AHEI, mean (SD) 49.4 (9.7) 45.1 (9.7) 49.9 (9.7) 46.7 (9.7) 50.2 (9.6) 44.3 (9.5)
EDIH, ≥ median 50.3 64.9 47.0 62.2 46.1 68.3
Total calories intake, kcal/day, mean (SD) 1711.0 (501.6) 1824.8 (557) 1699.1 (504.6) 1783.9 (526.4) 1691.5 (498.5) 1849.1 (559.5)
Regular use of aspirin, >2 tablets/week 28.1 27.2 27.8 28.1 27.8 27.4
Multivitamin use 48.3 41.9 49.5 43.5 49.9 40.6
Postmenopausal 89.2 88.3 89.4 88.2 89.5 88.0
Current postmenopausal hormone use 40.5 34.5 42.7 34.9 42.8 32.8
History of hypertension 23.7 21.5 21.2 26.0 21.9 23.3
History of hypercholesterolemia e 29.8 23.4 28.7 28.1 29.5 23.8
History of diabetes e 2.8 1.6 1.9 3.4 2.0 1.9
Family history of cancer 25.3 24.9 25.2 24.8 25.4 24.7
Family history of cardiovascular disease 7.7 7.5 7.5 7.7 7.6 7.7
History of depression 6.0 7.4 6.0 6.6 5.9 7.4
History of rotating night shiftwork 51.7 52.6 51.4 52.6 51.6 53.8
Husbands’ educational level
 Less than high school 1.4 2.1 1.3 1.8 1.3 2.3
 Some high school 3.0 3.7 2.9 3.6 2.8 4.2
 High school graduate 31.0 32.8 30.4 32.4 30.6 33.9
 College 24.0 21.6 23.9 23.1 24.2 21.4
 Graduate school 20.4 17.1 20.5 18.8 20.6 16.3
Meal frequency h
 1–2 times/day 3.7 5.8 4.2 4.0 4.0 5.5
 3–4 times/day 65.8 62.0 65.5 64.4 65.9 62.2
 5 or more times/day 30.5 32.2 30.3 31.7 30.1 32.3
Breakfast consumption h
 0–2 times/week 7.4 12.9 8.0 9.0 7.6 13.0
 3–5 times/week 9.4 13.1 9.7 10.9 9.3 13.5
 6–7 times/week 83.2 74.0 82.3 80.2 83.1 73.6
Open in a new tab

Abbreviations: NHS, Nurses’ Health Study; SD, standard deviation; BMI, body mass index; kg, kilogram; m, meter; cigs, cigarettes; g, gram; MET, metabolic equivalent task; AHEI, alternate healthy eating index; EDIH, empirical dietary index for hyperinsulinemia; NSAIDs, nonsteroidal anti-inflammatory drugs; kcal, kilocalorie, ug, microgram; mg, milligram; IU, international unit.

a

Data are expressed as percentages unless otherwise indicated.

b

Percentages are of non-missing values.

c

Percentages may not sum to 100% after rounding.

d

Calculated as weight in kilograms divided by height in meters squared.

e

Implementing a 4-year lag-time.

f

Cumulative among smokers.

g

Weekly energy expenditure in MET-hours/week from recreational and leisure time physical activity.

h

Assessed in 2002.

i

Defined as those exhibiting dietary restraint, i.e. those who reported both “not having the behavior of eating anything at any time” and “paying a great deal of attention to changes in their figure”.

Unrestrained Eating Behavior and Mortality Risk (Primary Analyses)

In multivariable-adjusted analyses, women with the behavior of eating anything at any time had an increased mortality from cancer (driven by gastrointestinal tract cancer) and respiratory disease, and decreased mortality from cardiovascular disease, compared to those without this behavior. Women who reported having no concern with figure change experienced higher risk of mortality from all-cause, cancer, and respiratory disease, compared to those not reporting this behavior. The combined effect of both eating anything at any time and having no concern with figure change was associated with a higher all-cause, cancer-specific (driven by cancers of gastrointestinal tract and lung), and respiratory disease-specific mortality, and was inversely associated with cardiovascular disease-specific mortality, compared to those exhibiting the opposite. Compared with age-adjusted analyses, the associations were attenuated after additionally adjusting for multiple covariates. Of note, controlling for EDIH attenuated the effect estimates for most of the aggregated and specific disease endpoints. Variations in the effect of adjustment for BMI were observed across exposures for lung cancer, breast cancer, hematopoietic cancer, and respiratory disease mortality. (Tables 2 and 3)

Table 2.

Hazard ratios (HR) and 95% confidence intervals (CI) of all-cause and cause-specific mortality according to self-reported unrestrained eating

NHS Eating anything at any time No concern with figure change Eating anything at any time + no concern with figure change
No Yes No Yes No c Yes
All-cause mortality
Cases 16440 5513 13395 8725 10394 2919
Age-adjusted a 1 [Reference] 1.26 (1.23–1.30) 1 1.29 (1.26–1.33) 1 1.46 (1.40–1.52)
Multivariable (without BMI) b 1 1.04 (1.01–1.07) 1 1.08 (1.05–1.11) 1 1.11 (1.06–1.16)
Multivariable (without EDIH) b 1 1.05 (1.02–1.08) 1 1.09 (1.06–1.12) 1 1.12 (1.07–1.17)
Multivariable b 1 1.02 (0.99–1.06) 1 1.08 (1.05–1.11) 1 1.09 (1.04–1.14)
Cancer mortality
Cases 4336 1581 3601 2353 2792 873
Age-adjusted a 1 1.33 (1.25–1.41) 1 1.25 (1.18–1.31) 1 1.53 (1.42–1.65)
Multivariable (without BMI) b 1 1.08 (1.01–1.14) 1 1.09 (1.03–1.15) 1 1.19 (1.10–1.29)
Multivariable (without EDIH) b 1 1.09 (1.02–1.16) 1 1.10 (1.04–1.16) 1 1.21 (1.11–1.31)
Multivariable b 1 1.07 (1.01–1.14) 1 1.08 (1.03–1.14) 1 1.19 (1.10–1.29)
Cardiovascular disease mortality
Cases 3493 1034 2756 1798 2158 518
Age-adjusted a 1 1.12 (1.04–1.20) 1 1.30 (1.23–1.38) 1 1.26 (1.14–1.39)
Multivariable (without BMI) b 1 0.94 (0.88–1.01) 1 1.05 (0.99–1.12) 1 0.94 (0.85–1.04)
Multivariable (without EDIH) b 1 0.94 (0.88–1.01) 1 1.03 (0.96–1.09) 1 0.90 (0.82–1.00)
Multivariable b 1 0.93 (0.86–1.00) 1 1.02 (0.95–1.08) 1 0.89 (0.80–0.98)
Respiratory disease mortality
Cases 1322 632 1093 823 832 342
Age-adjusted a 1 1.79 (1.62–1.97) 1 1.48 (1.35–1.62) 1 2.13 (1.87–2.42)
Multivariable (without BMI) b 1 1.24 (1.12–1.37) 1 1.14 (1.04–1.25) 1 1.33 (1.16–1.52)
Multivariable (without EDIH) b 1 1.23 (1.11–1.36) 1 1.22 (1.11–1.35) 1 1.40 (1.22–1.61)
Multivariable b 1 1.17 (1.06–1.29) 1 1.18 (1.07–1.30) 1 1.31 (1.14–1.51)
Open in a new tab

Abbreviations: HR, hazard ratio; CI, confidence interval; BMI, body mass index; EDIH, the empirical dietary index for hyperinsulinemia.

a

Models were controlled for age and follow-up cycle.

b

Models were controlled for age, follow-up cycle, race, cumulative average BMI (with 4-year lag time), BMI at age 18, smoking status, time since quit smoking, pack-years of smoking, alcohol intake, physical activity, total calories intake, EDIH, regular use of aspirin, multivitamin use, menopausal status, postmenopausal hormone use, history of hypertension, history of hypercholesterolemia (with 4-year lag time), history of diabetes (with 4-year lag time), family history of cancer, and family history of cardiovascular disease.

c

Defined as those exhibiting dietary restraint, i.e. those who reported both “not having the behavior of eating anything at any time” and “paying a great deal of attention to changes in their figure”.

Table 3.

Hazard ratios (HR) and 95% confidence intervals (CI) of cancer-specific mortality according to self-reported unrestrained eating

NHS Eating anything at any time No concern with figure change Eating anything at any time + no concern with figure change
No Yes No Yes No c Yes
Gastrointestinal tract cancer mortality
Cases 507 216 439 280 339 116
Age-adjusted a 1 [Reference] 1.54 (1.32–1.81) 1 1.21 (1.04–1.41) 1 1.64 (1.33–2.03)
Multivariable (without BMI) b 1 1.31 (1.11–1.54) 1 1.05 (0.90–1.23) 1 1.35 (1.08–1.69)
Multivariable (without EDIH) b 1 1.34 (1.14–1.58) 1 1.08 (0.93–1.27) 1 1.39 (1.11–1.74)
Multivariable b 1 1.31 (1.11–1.55) 1 1.06 (0.91–1.25) 1 1.36 (1.08–1.71)
 Colorectal cancer mortality
 Cases 356 138 301 194 240 79
 Age-adjusted a 1 1.41 (1.16–1.72) 1 1.23 (1.02–1.47) 1 1.59 (1.23–2.06)
 Multivariable (without BMI) b 1 1.22 (1.00–1.50) 1 1.10 (0.91–1.33) 1 1.40 (1.07–1.84)
 Multivariable (without EDIH) b 1 1.25 (1.02–1.53) 1 1.14 (0.94–1.38) 1 1.39 (1.06–1.82)
 Multivariable b 1 1.23 (1.00–1.51) 1 1.13 (0.93–1.36) 1 1.41 (1.07–1.86)
Digestive organ cancer mortality
Cases 528 151 402 271 327 82
Age-adjusted a 1 1.04 (0.87–1.25) 1 1.28 (1.10–1.50) 1 1.23 (0.96–1.57)
Multivariable (without BMI) b 1 0.92 (0.76–1.11) 1 1.12 (0.95–1.31) 1 0.99 (0.77–1.29)
Multivariable (without EDIH) b 1 0.93 (0.77–1.12) 1 1.09 (0.93–1.28) 1 0.97 (0.75–1.26)
Multivariable b 1 0.92 (0.76–1.11) 1 1.08 (0.92–1.27) 1 0.96 (0.74–1.24)
 Pancreatic cancer mortality
 Cases 371 107 287 188 231 55
 Age-adjusted a 1 1.06 (0.86–1.32) 1 1.26 (1.05–1.52) 1 1.18 (0.88–1.59)
 Multivariable (without BMI) b 1 0.92 (0.74–1.16) 1 1.10 (0.91–1.34) 1 0.96 (0.70–1.31)
 Multivariable (without EDIH) b 1 0.93 (0.75–1.16) 1 1.08 (0.89–1.32) 1 0.95 (0.69–1.29)
 Multivariable b 1 0.92 (0.74–1.15) 1 1.08 (0.88–1.31) 1 0.93 (0.68–1.28)
Lung cancer mortality
Cases 1006 466 844 613 643 259
Age-adjusted a 1 1.70 (1.52–1.89) 1 1.38 (1.25–1.54) 1 1.99 (1.72–2.30)
Multivariable (without BMI) b 1 1.13 (1.00–1.27) 1 1.13 (1.02–1.26) 1 1.28 (1.09–1.49)
Multivariable (without EDIH) b 1 1.13 (1.01–1.28) 1 1.21 (1.09–1.35) 1 1.35 (1.15–1.58)
Multivariable b 1 1.12 (1.00–1.26) 1 1.20 (1.08–1.34) 1 1.33 (1.14–1.56)
Breast cancer mortality
Cases 525 173 424 286 324 91
Age-adjusted a 1 1.18 (0.99–1.40) 1 1.28 (1.10–1.48) 1 1.33 (1.06–1.69)
Multivariable (without BMI) b 1 1.02 (0.85–1.21) 1 1.12 (0.96–1.31) 1 1.10 (0.86–1.40)
Multivariable (without EDIH) b 1 1.02 (0.86–1.22) 1 1.11 (0.95–1.30) 1 1.09 (0.85–1.40)
Multivariable b 1 1.02 (0.85–1.22) 1 1.10 (0.94–1.29) 1 1.10 (0.86–1.42)
Ovarian cancer mortality
Cases 305 114 272 156 206 60
Age-adjusted a 1 1.33 (1.07–1.65) 1 1.09 (0.90–1.33) 1 1.40 (1.05–1.87)
Multivariable (without BMI) b 1 1.16 (0.93–1.45) 1 0.99 (0.81–1.22) 1 1.26 (0.93–1.72)
Multivariable (without EDIH) b 1 1.19 (0.95–1.49) 1 1.01 (0.82–1.24) 1 1.29 (0.94–1.75)
Multivariable b 1 1.13 (0.90–1.42) 1 0.99 (0.80–1.21) 1 1.25 (0.91–1.70)
Hematopoietic cancer mortality
Cases 555 173 460 276 366 100
Age-adjusted a 1 1.13 (0.95–1.34) 1 1.15 (0.99–1.34) 1 1.34 (1.07–1.68)
Multivariable (without BMI) b 1 1.03 (0.86–1.23) 1 1.05 (0.90–1.22) 1 1.16 (0.91–1.46)
Multivariable (without EDIH) b 1 1.04 (0.87–1.24) 1 1.04 (0.89–1.22) 1 1.16 (0.92–1.48)
Multivariable b 1 1.04 (0.87–1.24) 1 1.03 (0.88–1.21) 1 1.17 (0.92–1.48)
Open in a new tab

Abbreviations: HR, hazard ratio; CI, confidence interval; BMI, body mass index; EDIH, the empirical dietary index for hyperinsulinemia.

a

Models were controlled for age and follow-up cycle.

b

Models were controlled for the same sets of covariates as denoted in Table 2.

c

Defined as those exhibiting dietary restraint, i.e. those who reported both “not having the behavior of eating anything at any time” and “paying a great deal of attention to changes in their figure”.

Unrestrained Eating Behavior and Mortality Risk (Sensitivity Analyses)

The robustness of these results was confirmed in sensitivity analyses. We performed latency analyses by excluding the first 4, 8, and 12 years of follow-up; however, this did not materially attenuate the observed associations for most of the endpoints, except for a suggestion of decreasing risk of respiratory disease mortality with longer lag-time. Replacing the EDIH with AHEI in confounding control, additionally controlling for history of depression, history of rotating night shiftwork, socioeconomic status, waist circumference, meal frequency, and breakfast consumption, and additionally adjusting for propensity score did not materially change our results for all the endpoints, with the only exception observed for lung cancer where slight to moderate attenuations of the effect estimates were observed after adjusting for propensity score. In sensitivity analyses restricted to participants reported ≤ 3 lifetime pack-years of smoking, the effect estimates remained pronounced. (Tables 4 and 5, and Supplementary Tables 1–8)

Table 4.

Hazard ratios (HR) and 95% confidence intervals (CI) of all-cause and cause-specific mortality according to self-reported unrestrained eating (sensitivity analyses: assuming 4-, 8- and 12-year latency)

NHS Eating anything at any time No concern with figure change Eating anything at any time + no concern with figure change
4-year latency (1998–2016) 8-year latency (2002–2016) 12-year latency (2006–2016) 4-year latency (1998–2016) 8-year latency (2002–2016) 12-year latency (2006–2016) 4-year latency (1998–2016) 8-year latency (2002–2016) 12-year latency (2006–2016)
Yes vs. No Yes vs. No Yes vs. No c
All-cause mortality
Cases 20735 18512 15184 20914 18706 15368 12596 11299 9325
Age-adjusted a 1.25 (1.21–1.29) 1.22 (1.18–1.27) 1.20 (1.15–1.24) 1.28 (1.25–1.32) 1.27 (1.24–1.31) 1.26 (1.22–1.30) 1.44 (1.38–1.50) 1.41 (1.34–1.47) 1.37 (1.30–1.44)
Multivariable (without BMI) b 1.03 (1.00–1.07) 1.02 (0.98–1.06) 1.01 (0.98–1.05) 1.08 (1.04–1.11) 1.08 (1.04–1.11) 1.07 (1.04–1.11) 1.09 (1.04–1.14) 1.09 (1.04–1.14) 1.08 (1.02–1.14)
Multivariable (without EDIH) b 1.04 (1.01–1.07) 1.02 (0.99–1.06) 1.02 (0.98–1.06) 1.09 (1.06–1.12) 1.08 (1.05–1.12) 1.08 (1.04–1.12) 1.11 (1.06–1.16) 1.10 (1.04–1.15) 1.09 (1.03–1.15)
Multivariable b 1.02 (0.98–1.05) 1.00 (0.97–1.04) 1.00 (0.96–1.04) 1.07 (1.04–1.10) 1.07 (1.04–1.10) 1.07 (1.03–1.11) 1.08 (1.03–1.13) 1.07 (1.02–1.12) 1.06 (1.01–1.12)
Cancer mortality
Cases 5346 4356 3057 5393 4397 3087 3318 2723 1936
Age-adjusted a 1.31 (1.23–1.39) 1.30 (1.21–1.39) 1.31 (1.20–1.42) 1.23 (1.17–1.30) 1.23 (1.16–1.31) 1.18 (1.10–1.27) 1.50 (1.39–1.63) 1.52 (1.39–1.67) 1.48 (1.33–1.65)
Multivariable (without BMI) b 1.07 (1.00–1.14) 1.07 (0.99–1.14) 1.09 (1.00–1.18) 1.08 (1.02–1.14) 1.10 (1.03–1.17) 1.06 (0.98–1.14) 1.18 (1.08–1.28) 1.22 (1.11–1.34) 1.21 (1.08–1.35)
Multivariable (without EDIH) b 1.08 (1.01–1.15) 1.07 (1.00–1.15) 1.10 (1.01–1.20) N.09 (1.03–1.15) 1.10 (1.03–1.17) 1.06 (0.98–1.14) 1.19 (1.09–1.29) 1.22 (1.11–1.34) 1.22 (1.08–1.36)
Multivariable b 1.06 (1.00–1.13) 1.07 (0.99–1.14) 1.09 (1.00–1.19) 1.08 (1.02–1.14) 1.09 (1.02–1.16) 1.05 (0.97–1.14) 1.17 (1.08–1.28) 1.22 (1.10–1.34) 1.21 (1.08–1.36)
Cardiovascular disease mortality
Cases 4253 3748 3065 4278 3778 3081 2526 2242 1848
Age-adjusted a 1.11 (1.04–1.20) 1.09 (1.01–1.18) 1.09 (1.00–1.19) 1.30 (1.22–1.38) 1.28 (1.20–1.37) 1.27 (1.18–1.37) 1.26 (1.14–1.39) 1.23 (1.10–1.37) 1.23 (1.09–1.38)
Multivariable (without BMI) b 0.95 (0.88–1.02) 0.93 (0.86–1.01) 0.95 (0.87–1.04) 1.06 (1.00–1.13) 1.06 (0.99–1.14) 1.07 (0.99–1.15) 0.95 (0.86–1.05) 0.95 (0.85–1.06) 0.97 (0.85–1.09)
Multivariable (without EDIH) b 0.95 (0.88–1.02) 0.93 (0.86–1.01) 0.95 (0.87–1.04) 1.03 (0.97–1.10) 1.03 (0.96–1.10) 1.04 (0.96–1.12) 0.91 (0.82–1.02) 0.91 (0.81–1.01) 0.94 (0.83–1.06)
Multivariable b 0.93 (0.86–1.00) 0.92 (0.85–0.99) 0.94 (0.86–1.03) 1.02 (0.96–1.09) 1.02 (0.95–1.09) 1.03 (0.95–1.11) 0.89 (0.80–0.99) 0.89 (0.79–1.00) 0.92 (0.81–1.04)
Respiratory disease mortality
Cases 1832 1621 1310 1802 1607 1310 1107 987 800
Age-adjusted a 1.74 (1.57–1.92) 1.64 (1.47–1.82) 1.52 (1.34–1.71) 1.45 (1.32–1.60) 1.39 (1.26–1.54) 1.35 (1.20–1.50) 2.05 (1.80–2.35) 1.91 (1.66–2.21) 1.75 (1.49–2.06)
Multivariable (without BMI) b 1.21 (1.09–1.34) 1.16 (1.04–1.30) 1.11 (0.97–1.26) 1.13 (1.03–1.25) 1.10 (0.99–1.22) 1.08 (0.96–1.21) 1.30 (1.13–1.49) 1.23 (1.05–1.43) 1.16 (0.98–1.38)
Multivariable (without EDIH) b 1.20 (1.08–1.33) 1.15 (1.03–1.29) 1.10 (0.97–1.25) 1.20 (1.09–1.33) 1.16 (1.04–1.29) 1.13 (1.01–1.27) 1.36 (1.17–1.57) 1.27 (1.08–1.48) 1.20 (1.01–1.43)
Multivariable b 1.14 (1.03–1.27) 1.10 (0.98–1.23) 1.06 (0.93–1.20) 1.17 (1.06–1.29) 1.13 (1.01–1.25) 1.11 (0.98–1.24) 1.28 (1.11–1.48) 1.20 (1.03–1.41) 1.15 (0.97–1.38)
Gastrointestinal tract cancer mortality
Cases 656 535 355 657 526 353 411 334 227
Age-adjusted a 1.49 (1.26–1.76) 1.55 (1.29–1.87) 1.56 (1.24–1.97) 1.20 (1.03–1.41) 1.23 (1.03–1.47) 1.27 (1.02–1.57) 1.59 (1.27–1.99) 1.67 (1.30–2.14) 1.77 (1.32–2.39)
Multivariable (without BMI) b 1.27 (1.07–1.52) 1.34 (1.11–1.63) 1.41 (1.11–1.78) 1.05 (0.89–1.23) 1.07 (0.89–1.28) 1.11 (0.89–1.39) 1.31 (1.03–1.66) 1.37 (1.05–1.78) 1.50 (1.09–2.05)
Multivariable (without EDIH) b 1.30 (1.09–1.55) 1.39 (1.14–1.68) 1.44 (1.14–1.83) 1.08 (0.91–1.27) 1.10 (0.91–1.32) 1.16 (0.92–1.45) 1.32 (1.04–1.68) 1.41 (1.09–1.84) 1.54 (1.12–2.11)
Multivariable b 1.27 (1.07–1.52) 1.35 (1.12–1.64) 1.43 (1.13–1.81) 1.06 (0.89–1.25) 1.08 (0.90–1.30) 1.14 (0.91–1.43) 1.30 (1.02–1.66) 1.38 (1.06–1.81) 1.53 (1.11–2.11)
Colorectal cancer mortality
Cases 447 367 243 452 360 241 285 233 157
Age-adjusted a 1.32 (1.07–1.63) 1.41 (1.12–1.77) 1.35 (1.02–1.80) 1.22 (1.01–1.48) 1.25 (1.01–1.54) 1.28 (0.99–1.66) 1.49 (1.13–1.97) 1.56 (1.15–2.11) 1.60 (1.10–2.31)
Multivariable (without BMI) b 1.14 (0.92–1.42) 1.23 (0.97–1.56) 1.21 (0.91–1.63) 1.08 (0.89–1.32) 1.10 (0.89–1.38) 1.14 (0.87–1.49) 1.30 (0.97–1.74) 1.36 (0.99–1.87) 1.40 (0.95–2.07)
Multivariable (without EDIH) b 1.16 (0.93–1.44) 1.26 (1.00–1.60) 1.25 (0.94–1.68) 1.11 (0.91–1.36) 1.16 (0.92–1.44) 1.20 (0.92–1.58) 1.26 (0.94–1.69) 1.39 (1.00–1.91) 1.43 (0.96–2.12)
Multivariable b 1.14 (0.92–1.42) 1.24 (0.98–1.58) 1.24 (0.92–1.66) 1.10 (0.90–1.35) 1.14 (0.91–1.43) 1.19 (0.91–1.57) 1.28 (0.95–1.72) 1.40 (1.01–1.94) 1.45 (0.97–2.15)
Lung cancer mortality
Cases 1298 1063 736 1285 1047 737 805 667 470
Age-adjusted a 1.67 (1.49–1.88) 1.71 (1.51–1.95) 1.72 (1.47–2.01) 1.34 (1.20–1.50) 1.41 (1.25–1.60) 1.32 (1.14–1.53) 1.92 (1.65–2.25) 2.08 (1.75–2.46) 2.02 (1.65–2.47)
Multivariable (without BMI) b 1.12 (0.99–1.27) 1.14 (0.99–1.30) 1.16 (0.99–1.37) 1.11 (0.99–1.25) 1.19 (1.05–1.35) 1.13 (0.97–1.32) 1.25 (1.06–1.48) 1.36 (1.13–1.63) 1.36 (1.09–1.70)
Multivariable (without EDIH) b 1.12 (0.99–1.27) 1.14 (0.99–1.31) 1.18 (1.00–1.40) 1.18 (1.05–1.33) 1.26 (1.10–1.43) 1.20 (1.02–1.40) 1.31 (1.11–1.55) 1.41 (1.18–1.70) 1.44 (1.16–1.80)
Multivariable b 1.11 (0.98–1.26) 1.14 (0.99–1.30) 1.17 (0.99–1.38) 1.18 (1.04–1.32) 1.25 (1.10–1.43) 1.19 (1.02–1.39) 1.30 (1.10–1.55) 1.41 (1.17–1.70) 1.42 (1.14–1.78)
Open in a new tab

Abbreviations: HR, hazard ratio; CI, confidence interval; BMI, body mass index; EDIH, the empirical dietary index for hyperinsulinemia.

a

Models were controlled for age and follow-up cycle.

b

Models were controlled for the same sets of covariates as denoted in Table 2.

c

Defined as those exhibiting dietary restraint, i.e. those who reported both “not having the behavior of eating anything at any time” and “paying a great deal of attention to changes in their figure”.

Table 5.

Hazard ratios (HR) and 95% confidence intervals (CI) of all-cause and cause-specific mortality according to self-reported unrestrained eating (sensitivity analyses: further adjusted for propensity score)

NHS Eating anything at any time No concern with figure change Eating anything at any time + no concern with figure change
No Yes No Yes No c Yes
All-cause mortality
Cases 16440 5513 13395 8725 10394 2919
Multivariable 1 a 1 [Reference] 1.02 (0.99–1.06) 1 1.08 (1.05–1.11) 1 1.09 (1.04–1.14)
Multivariable 2 b 1 1.02 (0.99–1.05) 1 1.08 (1.05–1.11) 1 1.09 (1.04–1.14)
Cancer mortality
Cases 4336 1581 3601 2353 2792 873
Multivariable 1 a 1 1.07 (1.01–1.14) 1 1.08 (1.03–1.15) 1 1.19 (1.10–1.29)
Multivariable 2 b 1 1.07 (1.00–1.13) 1 1.08 (1.02–1.14) 1 1.18 (1.09–1.28)
Cardiovascular disease mortality
Cases 3493 1034 2756 1798 2158 518
Multivariable 1 a 1 0.93 (0.86–1.00) 1 1.02 (0.96–1.08) 1 0.89 (0.80–0.98)
Multivariable 2 b 1 0.92 (0.86–0.99) 1 1.02 (0.96–1.08) 1 0.88 (0.80–0.98)
Respiratory disease mortality
Cases 1322 632 1093 823 832 342
Multivariable 1 a 1 1.17 (1.06–1.30) 1 1.19 (1.08–1.31) 1 1.32 (1.15–1.52)
Multivariable 2 b 1 1.16 (1.05–1.29) 1 1.18 (1.08–1.30) 1 1.30 (1.13–1.50)
Gastrointestinal tract cancer mortality
Cases 507 216 439 280 339 116
Multivariable 1 a 1 1.31 (1.11–1.55) 1 1.06 (0.91–1.25) 1 1.36 (1.08–1.70)
Multivariable 2 b 1 1.30 (1.10–1.54) 1 1.06 (0.91–1.24) 1 1.36 (1.08–1.71)
Colorectal cancer mortality
Cases 356 138 301 194 240 79
Multivariable 1 a 1 1.23 (1.00–1.51) 1 1.12 (0.93–1.36) 1 1.40 (1.07–1.85)
Multivariable 2 b 1 1.23 (1.00–1.51) 1 1.12 (0.93–1.36) 1 1.41 (1.07–1.86)
Lung cancer mortality
Cases 1006 466 844 613 643 259
Multivariable 1 a 1 1.12 (1.00–1.26) 1 1.20 (1.08–1.34) 1 1.34 (1.14–1.57)
Multivariable 2 b 1 1.02 (0.99–1.05) 1 1.08 (1.05–1.11) 1 1.09 (1.04–1.14)
Open in a new tab

Abbreviations: HR, hazard ratio; CI, confidence interval.

a

Models were controlled for the same sets of covariates as denoted in Table 2.

b

Models were controlled for the same sets of covariates as denoted in Table 2, and were further adjusted for propensity score.

c

Defined as those exhibiting dietary restraint, i.e. those who reported both “not having the behavior of eating anything at any time” and “paying a great deal of attention to changes in their figure”.

Discussion

In this large prospective investigation, unrestrained eating behaviors were associated with increased risk of all-cause mortality. When exploring contributors to these relationships, the risk elevation was predominately driven by mortalities from cancer (particularly for cancers of gastrointestinal tract and lung) and respiratory disease, rather than cardiovascular disease. On the contrary, unrestrained eating behaviors were associated with decreased risk of cardiovascular disease-specific mortality.

Comparison with Other Studies

To our knowledge, no previous study has been able to investigate unrestrained eating behavior in relation to all-cause or cause-specific mortality, though very few surrogate eating habits (eating frequency, snacking) have been explored by two investigations, with conflicting findings reported.[8, 9] In the National Health and Nutrition Examination Survey (NHANES) III, an inverse association was found between eating frequency and all-cause (suggestive) and cardiovascular disease mortality.[8] In another smaller study, snacking between meals was not associated with total mortality.[9] The associations of unrestrained eating and a wide range of related surrogate eating habits (e.g., overeating, frequent eating, mistimed/irregular meals, and snacking, etc.) with risk of other major health outcomes have been explored before; with mixed findings reported. Risk of cancer and cardiovascular disease are most frequently investigated in relation to eating behavior. A prior analysis from the NHS reported the behavior of eating anything at any time to be associated with higher gastrointestinal tract cancer risk.[32] A meta-analysis, pooling data from 3 cohort studies and 12 case-control studies, reported a positive association between eating frequency and colorectal cancer risk among case-control studies, but not among cohort studies.[33] Interestingly, controversial findings have been observed when disentangling the effect from snacking frequency with frequency of regular meals.[34–37] The relationships between surrogate eating habits and risk of overall digestive system cancer,[38, 39] several specific gastrointestinal tract cancers,[40, 41] and breast cancer[42–46] have also been explored, though the existing evidence remains relatively sparse. With very few exceptions,[41, 44] most of the studies supported positive associations.[38–43, 45, 46] Several surrogate eating habits have also been linked with cardiometabolic health.[47, 48] In 2017, the American Heart Association (AHA), after critical review of prior evidence, indicated the potential implications of modifying the timing and frequency of eating (including both meals and snacks) for cardiovascular disease prevention in their scientific statement as a crucial step.[48] There has been no evidence regarding the association of unrestrained eating (or surrogate eating habits) with risk or mortality of lung cancer and respiratory disease.

Of note, the prior epidemiological evidence examining surrogate eating habits may have been limited by their study design (relatively small sample size, predominantly case-control studies, imprecise/incomplete assessment of eating behaviors, inadequate confounding control, and lack of consideration for potential reverse causation in data analyses and result interpretation), offering possible explanations of the significant heterogeneity detected in these studies and the discrepancy observed in their findings. Interventional studies have also been limited in their scope, and are too diverse to draw definitive conclusions or make recommendations.[48] Our present study, overcoming most of the above-mentioned limitations, contradicts the prior reports on all-cause mortality, but largely corroborates the previous evidence linking surrogate eating habits with risk of other major health outcomes. Nevertheless, it is not possible to compare our results with most of these studies directly, due to the variations in their exposures (unrestrained eating vs. surrogate eating habits), outcomes (mortality vs. incidence), and study methods. In addition, given the NHS participants have previously been suggested to have healthier lifestyles than the general population,[2] the possibility that there could be differences in the magnitudes of the associations when assessing the general US population cannot be ruled out. Interestingly, although the second question “I pay a great deal of attention to changes in my figure” is part of the “flexible, but not rigid control’ subscale proposed for dietary restraint, which is considered to have greater benefits for counterbalancing dietary disinhibition than other aspects of dietary restraint,[22] we observed similar effects in analyses across all three exposures.

A higher mortality risk from all-cause, cancers of gastrointestinal tract and lung, and respiratory disease observed among unrestrained eaters is biologically plausible. Disturbed food intake rhythms among unrestrained eaters may cause circadian disruption,[42, 49–51] and affect multiple glucose homeostasis regulation (e.g., insulin sensitivity), energy metabolism, and immune function,[49, 52–54] which have been associated with higher all-cause and cause-specific mortality.[55–60] Moreover, it may operate in conjunction with other altered lifestyle behaviors that are more prevalent among unrestrained eaters (e.g., late-night eating, large meals, fatty or fried foods, alcohol and tobacco use, obesity, etc.), further increase the chance of mucus-lined barrier injury, chronic inflammation, and malignant transformation in the gastrointestinal tract. The mechanisms specifically driving the observed association of unrestrained eating with mortality from respiratory disease and lung cancer remain to be elucidated, with insulin resistance having been suggested by prior investigations,[61–70] including evidence from Mendelian randomization studies.[69, 70] It is noteworthy that in our multivariable analyses, controlling for EDIH (an empirical index developed to specifically assess the insulinemic potential of usual diets, comprising 18 food groups previously identified to be most predictive of fasting plasma C-peptide in stepwise regression),[71] attenuated the effect estimates for most of the aggregated and specific disease endpoints, including mortality from all-cause, gastrointestinal tract cancer, lung cancer, and respiratory disease. On the other hand, replacing EDIH with AHEI-2010 (which has been strongly associated with major chronic disease risk, including diabetes)[72] did not materially change the effect estimates for all the above-mentioned endpoints. Our results support the potential role of prolonged hyperinsulinemia in the putative mechanisms linking unrestrained eating with mortality risk.

The observation of a statistically significant lower risk of cardiovascular disease-specific mortality associated with unrestrained eaters in analyses of the combined effect of both eating anything at any time and having no concern with figure change is unexpected. The AHA statement generally suggests the benefit of daily breakfast consumption in promoting cardiometabolic health, [48] whereas evidence regarding the effect of eating frequency remains inconclusive.[48] In the only large prospective cohort study that quantified the relationship between eating frequency and risk of coronary heart disease, no association was observed after rigorous confounding control.[73] Interestingly, in the NHANES III, eating frequency was inversely associated with cardiovascular disease-specific mortality.[8] In our sensitivity analyses, however, additionally controlling for eating frequency and breakfast skipping did not change the effect estimates for mortality from cardiovascular disease. (Supplementary Tables 3–5) Of note, in all the combined effect analyses, we specifically excluded participants who “ate anything at any time but cared about their figure change” and those who “did not eat anything at any time but did not care about their figure change” to have a clear contrast, and we used those exhibiting dietary restraint as the reference group. Until this finding can be confirmed in other large epidemiologic or interventional studies and the underlying causes of the association could be identified, it requires cautious interpretation.

Strengths and Limitations of Study

Our study has several major strengths. First, the prospective cohort study design, large sample size, high follow-up rate, large number of deaths documented during long follow-up, in conjunction with a standard review of death and medical records by investigators, offers a high-quality dataset to investigate this topic. Second, repeatedly assessed and extensively validated information on anthropometrics, lifestyle, diet, and medical and family histories allowed for rigorous confounding adjustment.[10–12, 14, 15, 23, 24, 31] In fact, methods developed from the NHS for reducing bias attributable to measurement error and misclassification have been intensively applied to other epidemiological studies, and have been increasingly applied to implementation science, comparative effectiveness research, and the evaluation of public health interventions.[31] Third, the nature of our study participants (all health professionals) further ensured data quality, enhanced internal validity, and reduced socioeconomic confounding. Lastly, we were able to perform comprehensive sensitivity analyses to test the robustness of our findings.

Limitations of this study also deserve comment. First, we were not able to cumulatively update exposure assessment during follow-up. However, given the effect estimates were generally not attenuated in latency analyses, potential concerns about reverse causation and whether these assessments may reflect a long-term eating behavior status have been alleviated. Of note, although a validation study of measurements of unrestrained eating has not been performed, we have similar confidence in the reliability of our exposure assessments given a wide range of diet and lifestyle measurements have previously been demonstrated to be highly valid in the NHS cohort. Further, the fact that exposures were measured once at baseline could have attenuated the effect estimates towards the null. Nevertheless, future investigations with more refined assessments of exposure are warranted. Second, despite the extensive control for all available confounding factors, residual and unmeasured confounding cannot be ruled out. Especially, our exposure of interest may be both a behavior and an attitude (Table 1: unrestrained eaters generally have less health conscious behaviors), and the possibility of bias in dietary reporting by restrained versus unrestrained eaters[74] may have added further complexity. However, we performed the propensity score analyses, and results remained largely robust. Lastly, restricting the study sample to predominantly white female healthcare professionals aged 48–73 years in 1994 when information on unrestrained eating behaviors were first assessed may limit the generalizability of our findings.

Conclusion

This large prospective investigation supports the associations between unrestrained eating and increased risk of all-cause, cancer-specific (particularly for gastrointestinal tract cancer and lung cancer), and respiratory disease-specific mortality. These findings highlight the importance of unrestrained eating behavior modification, independent of major lifestyle, diet, anthropometric, and demographic predictors of mortality, in preventing death. The inverse association observed between unrestrained eating and cardiovascular disease-specific mortality should be interpreted with caution, and warrants further exploration and validation.

Supplementary Material

1

Acknowledgements

The authors thank all participants and staff of the Nurses’ Health Studies for their great contributions to this research. We are grateful for help from the following state cancer registries: Alabama, Arizona, Arkansas, California, Colorado, Connecticut, Delaware, Florida, Georgia, Idaho, Illinois, Indiana, Iowa, Kentucky, Louisiana, Maine, Maryland, Massachusetts, Michigan, Nebraska, New Hampshire, New Jersey, New York, North Carolina, North Dakota, Ohio, Oklahoma, Oregon, Pennsylvania, Rhode Island, South Carolina, Tennessee, Texas, Virginia, Washington, and Wyoming. The authors assume full responsibility for analyses and interpretation of these data.

Funding Statement

This work was supported by the National Institutes of Health grants UM1 CA186107 and P01 CA87969 to the Nurses’ Health Study. This work was additionally supported by the American Cancer Society Mentored Research Scholar Grant MRSG-17-220-01-NEC and the National Institutes of Health grant R00 CA215314 to MS; by Pussycat Foundation Helen Gurley Brown Presidential Initiative to CY; by National Institutes of Health grants R01 CA137178, and R35 CA253185 to ATC; ATC is also a Stuart and Suzanne Steele MGH Research Scholar; by National Institutes of Health grant R01 OH009803 to ESS; by Lustgarten Foundation and Dana-Farber Cancer Institute Hale Family Center for Pancreatic Cancer Research, National Institutes of Health grant U01 CA210171, Pancreatic Cancer Action Network, Stand Up to Cancer, Noble Effort Fund, Wexler Family Fund, and Promises for Purple to BMW; by National Institutes of Health grant P50 CA127003 to CSF; by a grant from the World Cancer Research Fund to ELG; by National Institutes of Health grant R01 CA205406, Department of Defense grant CA160344, Project P Fund, Broman Family Fund, and Pussycat Foundation Helen Gurley Brown Presidential Initiative to KN; and by the Entertainment Industry Foundation’s National Colorectal Cancer Research Alliance (NCCRA). This research was also supported by Stand Up To Cancer-Lustgarten Foundation Pancreatic Cancer Interception Translational Cancer Research grant SU2C-AACR-DT25-17. The funding sources played no role in the study design, data collection, data analysis, and interpretation of results, or the decisions made in preparation and submission of the article.

Conflicts of interest

BMW declares research funding from Celgene and Eli Lilly and Company and consulting for BioLineRx, Celgene, G1 Therapeutics, and GRAIL, outside the submitted work. ATC declares research funding from Bayer and consulting for Bayer, Pfizer Inc. and Boehringer Ingelheim, outside the submitted work. JAM declares institutional research funding from Boston Biomedical and consulting for Ignyta, Taiho Pharmaceutical, and Cota, outside the submitted work. CSF declares consulting for Agios, Bain Capital, Bayer, Celgene, Dicerna Pharmaceuticals, Eli Lilly and Company, Entrinsic Health Solutions, Five Prime Therapeutics, Genentech, Gilead Sciences, KEW, Merck & Co., Merrimack Pharmaceuticals, Pfizer, Sanofi, Taiho Pharmaceutical, and Unum Therapeutics, outside the submitted work. He also serves as a Director for CytomX Therapeutics and owns unexercised stock options for CytomX Therapeutics and Entrinsic Health Solutions, outside the submitted work. KN declares institutional research funding from Revolution Medicines, Celgene, Genentech, Gilead Sciences, Pharmavite, Tarrex Biopharma, and Trovagene, and consulting/advisory board fees from Array Biopharma, Bayer, Eli Lilly, Genentech, Seattle Genetics, and Tarrex Biopharma, outside the submitted work. Other authors declare no potential conflicts of interest.

Footnotes

Ethics approval

The study protocol was approved by the Institutional Review Board of the Brigham and Women’s Hospital (Boston, MA), and those of participating registries as required. Informed consent from participants was implied by the completion and return of the questionnaires. Written informed consent was required to retrieve medical records.

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References:

  • [1].Islami F, Goding Sauer A, Miller KD, Siegel RL, Fedewa SA, Jacobs EJ, et al. Proportion and number of cancer cases and deaths attributable to potentially modifiable risk factors in the United States. CA Cancer J Clin. 2018;68:31–54. [DOI] [PubMed] [Google Scholar]
  • [2].Song M, Giovannucci E. Preventable Incidence and Mortality of Carcinoma Associated With Lifestyle Factors Among White Adults in the United States. JAMA Oncol. 2016;2:1154–61. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [3].Li Y, Schoufour J, Wang DD, Dhana K, Pan A, Liu X, et al. Healthy lifestyle and life expectancy free of cancer, cardiovascular disease, and type 2 diabetes: prospective cohort study. BMJ. 2020;368:l6669. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [4].Veronese N, Li Y, Manson JE, Willett WC, Fontana L, Hu FB. Combined associations of body weight and lifestyle factors with all cause and cause specific mortality in men and women: prospective cohort study. BMJ. 2016;355:i5855. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [5].Li Y, Pan A, Wang DD, Liu X, Dhana K, Franco OH, et al. Impact of Healthy Lifestyle Factors on Life Expectancies in the US Population. Circulation. 2018;138:345–55. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [6].Kerr J, Anderson C, Lippman SM. Physical activity, sedentary behaviour, diet, and cancer: an update and emerging new evidence. Lancet Oncol. 2017;18:e457–e71. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [7].Sotos-Prieto M, Bhupathiraju SN, Mattei J, Fung TT, Li Y, Pan A, et al. Association of Changes in Diet Quality with Total and Cause-Specific Mortality. N Engl J Med. 2017;377:143–53. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [8].Chen HJ, Wang Y, Cheskin LJ. Relationship between frequency of eating and cardiovascular disease mortality in U.S. adults: the NHANES III follow-up study. Ann Epidemiol. 2016;26:527–33. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [9].Wingard DL, Berkman LF, Brand RJ. A multivariate analysis of health-related practices: a nine-year mortality follow-up of the Alameda County Study. Am J Epidemiol. 1982;116:765–75. [DOI] [PubMed] [Google Scholar]
  • [10].Wolf AM, Hunter DJ, Colditz GA, Manson JE, Stampfer MJ, Corsano KA, et al. Reproducibility and validity of a self-administered physical activity questionnaire. Int J Epidemiol. 1994;23:991–9. [DOI] [PubMed] [Google Scholar]
  • [11].Rimm EB, Stampfer MJ, Colditz GA, Chute CG, Litin LB, Willett WC. Validity of self-reported waist and hip circumferences in men and women. Epidemiology. 1990;1:466–73. [DOI] [PubMed] [Google Scholar]
  • [12].Feskanich D, Rimm EB, Giovannucci EL, Colditz GA, Stampfer MJ, Litin LB, et al. Reproducibility and validity of food intake measurements from a semiquantitative food frequency questionnaire. J Am Diet Assoc. 1993;93:790–6. [DOI] [PubMed] [Google Scholar]
  • [13].Hu FB, Rimm E, Smith-Warner SA, Feskanich D, Stampfer MJ, Ascherio A, et al. Reproducibility and validity of dietary patterns assessed with a food-frequency questionnaire. Am J Clin Nutr. 1999;69:243–9. [DOI] [PubMed] [Google Scholar]
  • [14].Colditz GA, Martin P, Stampfer MJ, Willett WC, Sampson L, Rosner B, et al. Validation of questionnaire information on risk factors and disease outcomes in a prospective cohort study of women. Am J Epidemiol. 1986;123:894–900. [DOI] [PubMed] [Google Scholar]
  • [15].Colditz GA, Stampfer MJ, Willett WC, Stason WB, Rosner B, Hennekens CH, et al. Reproducibility and validity of self-reported menopausal status in a prospective cohort study. Am J Epidemiol. 1987;126:319–25. [DOI] [PubMed] [Google Scholar]
  • [16].Belanger CF, Hennekens CH, Rosner B, Speizer FE. The nurses’ health study. Am J Nurs. 1978;78:1039–40. [PubMed] [Google Scholar]
  • [17].Belanger C, Speizer FE, Hennekens CH, Rosner B, Willett W, Bain C. The nurses’ health study: current findings. Am J Nurs. 1980;80:1333. [DOI] [PubMed] [Google Scholar]
  • [18].Colditz GA, Manson JE, Hankinson SE. The Nurses’ Health Study: 20-year contribution to the understanding of health among women. J Womens Health. 1997;6:49–62. [DOI] [PubMed] [Google Scholar]
  • [19].Colditz GA, Hankinson SE. The Nurses’ Health Study: lifestyle and health among women. Nat Rev Cancer. 2005;5:388–96. [DOI] [PubMed] [Google Scholar]
  • [20].Stunkard AJ, Messick S. The three-factor eating questionnaire to measure dietary restraint, disinhibition and hunger. J Psychosom Res. 1985;29:71–83. [DOI] [PubMed] [Google Scholar]
  • [21].Bao Y, Nimptsch K, Chan AT, Ng K, Meyerhardt JA, Willett WC, et al. Reported behavior of eating anything at anytime and risk of colorectal cancer in women. Int J Cancer. 2012;130:1395–400. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [22].Hays NP, Roberts SB. Aspects of eating behaviors “disinhibition” and “restraint” are related to weight gain and BMI in women. Obesity (Silver Spring). 2008;16:52–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [23].Stampfer MJ, Willett WC, Speizer FE, Dysert DC, Lipnick R, Rosner B, et al. Test of the National Death Index. Am J Epidemiol. 1984;119:837–9. [DOI] [PubMed] [Google Scholar]
  • [24].Rich-Edwards JW, Corsano KA, Stampfer MJ. Test of the National Death Index and Equifax Nationwide Death Search. Am J Epidemiol. 1994;140:1016–9. [DOI] [PubMed] [Google Scholar]
  • [25].Brazier JE, Harper R, Jones NM, O’Cathain A, Thomas KJ, Usherwood T, et al. Validating the SF-36 health survey questionnaire: new outcome measure for primary care. BMJ (Clinical research ed). 1992;305:160–4. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [26].McCullough ML, Willett WC. Evaluating adherence to recommended diets in adults: the Alternate Healthy Eating Index. Public Health Nutr. 2006;9:152–7. [DOI] [PubMed] [Google Scholar]
  • [27].Wang W, Fung TT, Wang M, Smith-Warner SA, Giovannucci EL, Tabung FK. Association of the Insulinemic Potential of Diet and Lifestyle With Risk of Digestive System Cancers in Men and Women. JNCI Cancer Spectr. 2018;2:pky080. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [28].Tabung FK, Wang W, Fung TT, Smith-Warner SA, Keum N, Wu K, et al. Association of dietary insulinemic potential and colorectal cancer risk in men and women. Am J Clin Nutr. 2018;108:363–70. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [29].Berwick DM, Murphy JM, Goldman PA, Ware JE Jr., Barsky AJ, Weinstein MC. Performance of a five-item mental health screening test. Medical care. 1991;29:169–76. [DOI] [PubMed] [Google Scholar]
  • [30].Weinstein MC, Berwick DM, Goldman PA, Murphy JM, Barsky AJ. A comparison of three psychiatric screening tests using receiver operating characteristic (ROC) analysis. Medical care. 1989;27:593–607. [DOI] [PubMed] [Google Scholar]
  • [31].Spiegelman D. Evaluating Public Health Interventions: 4. The Nurses’ Health Study and Methods for Eliminating Bias Attributable to Measurement Error and Misclassification. Am J Public Health. 2016;106:1563–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [32].Zhang Y, Song M, Chan AT, Schernhammer ES, Wolpin BM, Stampfer MJ, et al. Unrestrained eating behavior and risk of digestive system cancers: a prospective cohort study. Am J Clin Nutr. 2021. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [33].Liu Y, Tang W, Zhai L, Yang S, Wu J, Xie L, et al. Meta-analysis: eating frequency and risk of colorectal cancer. Tumour Biol. 2014;35:3617–25. [DOI] [PubMed] [Google Scholar]
  • [34].de Verdier MG, Longnecker MP. Eating frequency--a neglected risk factor for colon cancer? Cancer Causes Control. 1992;3:77–81. [DOI] [PubMed] [Google Scholar]
  • [35].Tseng M, Ingram DD, Darden R, Ziegler RG, Longnecker MP. Eating frequency and risk of colorectal cancer. Nutr Cancer. 2000;36:170–6. [DOI] [PubMed] [Google Scholar]
  • [36].Shoff SM, Newcomb PA, Longnecker MP. Frequency of eating and risk of colorectal cancer in women. Nutr Cancer. 1997;27:22–5. [DOI] [PubMed] [Google Scholar]
  • [37].Perrigue MM, Kantor ED, Hastert TA, Patterson R, Potter JD, Neuhouser ML, et al. Eating frequency and risk of colorectal cancer. Cancer Causes Control. 2013;24:2107–15. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [38].Chen P, Lin Y, Zheng K, Liu B, Wu C, Yan W, et al. Risk Factors of Gastric Cancer in High-Risk Region of China: A Population-Based Case-control Study. Asian Pac J Cancer Prev. 2019;20:775–81. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [39].Ye WM, Yi YN, Luo RX, Zhou TS, Lin RT, Chen GD. Diet and gastric cancer: a casecontrol study in Fujian Province, China. World J Gastroenterol. 1998;4:516–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [40].Cai L, Zheng ZL, Zhang ZF. Risk factors for the gastric cardia cancer: a case-control study in Fujian Province. World J Gastroenterol. 2003;9:214–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [41].Ibiebele TI, Taylor AR, Whiteman DC, van der Pols JC, Australian Cancer S. Eating habits and risk of esophageal cancers: a population-based case-control study. Cancer Causes Control. 2010;21:1475–84. [DOI] [PubMed] [Google Scholar]
  • [42].Srour B, Plancoulaine S, Andreeva VA, Fassier P, Julia C, Galan P, et al. Circadian nutritional behaviours and cancer risk: New insights from the NutriNet-sante prospective cohort study: Disclaimers. Int J Cancer. 2018;143:2369–79. [DOI] [PubMed] [Google Scholar]
  • [43].Kogevinas M, Espinosa A, Castello A, Gomez-Acebo I, Guevara M, Martin V, et al. Effect of mistimed eating patterns on breast and prostate cancer risk (MCC-Spain Study). Int J Cancer. 2018;143:2380–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [44].Marinac CR, Sears DD, Natarajan L, Gallo LC, Breen CI, Patterson RE. Frequency and Circadian Timing of Eating May Influence Biomarkers of Inflammation and Insulin Resistance Associated with Breast Cancer Risk. PLoS One. 2015;10:e0136240. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [45].Li M, Tse LA, Chan WC, Kwok CH, Leung SL, Wu C, et al. Nighttime eating and breast cancer among Chinese women in Hong Kong. Breast Cancer Res. 2017;19:31. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [46].Kim JH, Lee J, Jung SY, Kim J. Dietary Factors and Female Breast Cancer Risk: A Prospective Cohort Study. Nutrients. 2017;9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [47].Maugeri A, Vinciguerra M. The Effects of Meal Timing and Frequency, Caloric Restriction, and Fasting on Cardiovascular Health: an Overview. J Lipid Atheroscler. 2020;9:140–52. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [48].St-Onge MP, Ard J, Baskin ML, Chiuve SE, Johnson HM, Kris-Etherton P, et al. Meal Timing and Frequency: Implications for Cardiovascular Disease Prevention: A Scientific Statement From the American Heart Association. Circulation. 2017;135:e96–e121. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [49].Depner CM, Melanson EL, McHill AW, Wright KP Jr. Mistimed food intake and sleep alters 24-hour time-of-day patterns of the human plasma proteome. Proc Natl Acad Sci U S A. 2018;115:E5390–E9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [50].Salgado-Delgado R, Angeles-Castellanos M, Saderi N, Buijs RM, Escobar C. Food intake during the normal activity phase prevents obesity and circadian desynchrony in a rat model of night work. Endocrinology. 2010;151:1019–29. [DOI] [PubMed] [Google Scholar]
  • [51].Li XM, Delaunay F, Dulong S, Claustrat B, Zampera S, Fujii Y, et al. Cancer inhibition through circadian reprogramming of tumor transcriptome with meal timing. Cancer Res. 2010;70:3351–60. [DOI] [PubMed] [Google Scholar]
  • [52].Yoon JA, Han DH, Noh JY, Kim MH, Son GH, Kim K, et al. Meal time shift disturbs circadian rhythmicity along with metabolic and behavioral alterations in mice. PLoS One. 2012;7:e44053. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [53].Bishehsari F, Engen PA, Voigt RM, Swanson G, Shaikh M, Wilber S, et al. Abnormal Eating Patterns Cause Circadian Disruption and Promote Alcohol-Associated Colon Carcinogenesis. Cell Mol Gastroenterol Hepatol. 2020;9:219–37. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [54].Stenvers DJ, Scheer F, Schrauwen P, la Fleur SE, Kalsbeek A. Circadian clocks and insulin resistance. Nat Rev Endocrinol. 2019;15:75–89. [DOI] [PubMed] [Google Scholar]
  • [55].Lakka HM, Lakka TA, Tuomilehto J, Sivenius J, Salonen JT. Hyperinsulinemia and the risk of cardiovascular death and acute coronary and cerebrovascular events in men: the Kuopio Ischaemic Heart Disease Risk Factor Study. Arch Intern Med. 2000;160:1160–8. [DOI] [PubMed] [Google Scholar]
  • [56].Khaw KT, Wareham N, Luben R, Bingham S, Oakes S, Welch A, et al. Glycated haemoglobin, diabetes, and mortality in men in Norfolk cohort of european prospective investigation of cancer and nutrition (EPIC-Norfolk). BMJ. 2001;322:15–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [57].Barone BB, Yeh HC, Snyder CF, Peairs KS, Stein KB, Derr RL, et al. Long-term all-cause mortality in cancer patients with preexisting diabetes mellitus: a systematic review and meta-analysis. JAMA. 2008;300:2754–64. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [58].Lam EKK, Batty GD, Huxley RR, Martiniuk ALC, Barzi F, Lam TH, et al. Associations of diabetes mellitus with site-specific cancer mortality in the Asia-Pacific region. Ann Oncol. 2011;22:730–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [59].Tseng CH. Diabetes conveys a higher risk of gastric cancer mortality despite an age-standardised decreasing trend in the general population in Taiwan. Gut. 2011;60:774–9. [DOI] [PubMed] [Google Scholar]
  • [60].Wang M, Muraki I, Liu K, Shirai K, Tamakoshi A, Hu Y, et al. Diabetes and mortality from respiratory diseases: the Japan Collaborative Cohort Study. J Epidemiol. 2019. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [61].Machado FVC, Pitta F, Hernandes NA, Bertolini GL. Physiopathological relationship between chronic obstructive pulmonary disease and insulin resistance. Endocrine. 2018;61:17–22. [DOI] [PubMed] [Google Scholar]
  • [62].Piazzolla G, Castrovilli A, Liotino V, Vulpi MR, Fanelli M, Mazzocca A, et al. Metabolic syndrome and Chronic Obstructive Pulmonary Disease (COPD): The interplay among smoking, insulin resistance and vitamin D. PLoS One. 2017;12:e0186708. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [63].Argirion I, Weinstein SJ, Mannisto S, Albanes D, Mondul AM. Serum Insulin, Glucose, Indices of Insulin Resistance, and Risk of Lung Cancer. Cancer Epidemiol Biomarkers Prev. 2017;26:1519–24. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [64].Wiebe N, Stenvinkel P, Tonelli M. Associations of Chronic Inflammation, Insulin Resistance, and Severe Obesity With Mortality, Myocardial Infarction, Cancer, and Chronic Pulmonary Disease. JAMA Netw Open. 2019;2:e1910456. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [65].Cardet JC, Ash S, Kusa T, Camargo CA Jr., Israel E. Insulin resistance modifies the association between obesity and current asthma in adults. Eur Respir J. 2016;48:403–10. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [66].Petridou ET, Sergentanis TN, Antonopoulos CN, Dessypris N, Matsoukis IL, Aronis K, et al. Insulin resistance: an independent risk factor for lung cancer? Metabolism. 2011;60:1100–6. [DOI] [PubMed] [Google Scholar]
  • [67].Ehrlich SF, Quesenberry CP Jr., Van Den Eeden SK, Shan J, Ferrara A. Patients diagnosed with diabetes are at increased risk for asthma, chronic obstructive pulmonary disease, pulmonary fibrosis, and pneumonia but not lung cancer. Diabetes care. 2010;33:55–60. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [68].Fernandez-Real JM, Chico B, Shiratori M, Nara Y, Takahashi H, Ricart W. Circulating surfactant protein A (SP-A), a marker of lung injury, is associated with insulin resistance. Diabetes care. 2008;31:958–63. [DOI] [PubMed] [Google Scholar]
  • [69].Yuan S, Kar S, Carter P, Vithayathil M, Mason AM, Burgess S, et al. Is Type 2 Diabetes Causally Associated With Cancer Risk? Evidence From a Two-Sample Mendelian Randomization Study. Diabetes. 2020;69:1588–96. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [70].Carreras-Torres R, Johansson M, Haycock PC, Wade KH, Relton CL, Martin RM, et al. Obesity, metabolic factors and risk of different histological types of lung cancer: A Mendelian randomization study. PLoS One. 2017;12:e0177875. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [71].Tabung FK, Wang W, Fung TT, Hu FB, Smith-Warner SA, Chavarro JE, et al. Development and validation of empirical indices to assess the insulinaemic potential of diet and lifestyle. Br J Nutr. 2016:1–12. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [72].Chiuve SE, Fung TT, Rimm EB, Hu FB, McCullough ML, Wang M, et al. Alternative dietary indices both strongly predict risk of chronic disease. J Nutr. 2012;142:1009–18. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [73].Cahill LE, Chiuve SE, Mekary RA, Jensen MK, Flint AJ, Hu FB, et al. Prospective study of breakfast eating and incident coronary heart disease in a cohort of male US health professionals. Circulation. 2013;128:337–43. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [74].Bathalon GP, Tucker KL, Hays NP, Vinken AG, Greenberg AS, McCrory MA, et al. Psychological measures of eating behavior and the accuracy of 3 common dietary assessment methods in healthy postmenopausal women. Am J Clin Nutr. 2000;71:739–45. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

1
NIHMS1741204-supplement-1.docx (85KB, docx)

RESOURCES