Skip to main content
PMC home page
Cancer Science logoLink to Cancer Science
. 2024 Feb 4;115(4):1346–1359. doi: 10.1111/cas.16093

Body mass index and lung cancer risk: Pooled analysis of 10 prospective cohort studies in Japan

Sayo Kawai 1, Yingsong Lin 1,✉, Hiroshi Tsuge 2, Hidemi Ito 2,3, Keitaro Matsuo 4,5, Keiko Wada 6, Chisato Nagata 6, Nobuhiro Narii 7, Tetsuhisa Kitamura 7, Mai Utada 8, Ritsu Sakata 8, Takashi Kimura 9, Akiko Tamakoshi 9, Yumi Sugawara 10, Ichiro Tsuji 10, Seitaro Suzuki 11, Norie Sawada 11, Shoichiro Tsugane 11,12, Tetsuya Mizoue 13, Isao Oze 4, Sarah Krull Abe 14, Manami Inoue 14; Research Group for the Development and Evaluation of Cancer Prevention Strategies in Japan
PMCID: PMC11007012  PMID: 38310695

Abstract

Mounting evidence suggests that body mass index (BMI) is inversely associated with the risk of lung cancer. However, relatively few studies have explored this association in Asian people, who have a much lower prevalence of obesity than Caucasians. We pooled data from 10 prospective cohort studies involving 444,143 Japanese men and women to address the association between BMI and the risk of lung cancer. Study‐specific hazard ratios (HRs) and 95% confidence intervals (CIs) were calculated in each cohort using the Cox proportional hazards model. A meta‐analysis was undertaken by combining the results from each cohort. Heterogeneity across studies was evaluated using Cochran's Q and I 2 statistics. During 5,730,013 person‐years of follow‐up, 6454 incident lung cancer cases (4727 men and 1727 women) were identified. Baseline BMI was inversely associated with lung cancer risk in men and women combined. While leanness (BMI <18.5) was associated with a higher risk of lung cancer (HR 1.35; 95% CI, 1.16–1.57), overweight and obesity were associated with a lower risk, with HRs of 0.77 (95% CI, 0.71–0.84) and 0.69 (95% CI, 0.45–1.07), respectively. Every 5 kg/m2 increase in BMI was associated with a 21% lower risk of lung cancer (HR 0.79; 95% CI, 0.75–0.83; p < 0.0001). Our pooled analysis indicated that BMI is inversely associated with the risk of lung cancer in the Japanese population. This inverse association could be partly attributed to residual confounding by smoking, as it was more pronounced among male smokers.

Keywords: body mass index, inverse association, lung cancer risk, pooled analysis, smoking

Abbreviations

8‐OHdG

8‐hydroxy‐2‐deoxyguanosine

BMI

body mass index

CI

confidence interval

HR

hazard ratio

MR

Mendelian randomization

PY

pack year

1. INTRODUCTION

Obesity, as defined by BMI, is known to increase the risk of a wide range of diseases, including type 2 diabetes, cardiovascular disease, and cancer. 1 Over the past several decades, there has been growing interest in exploring the associations of BMI with cancer incidence and mortality. According to the 2018 working group report of the IARC, obesity has been shown to increase the risk of at least 13 different types of cancer. 2 However, one notable exception appears to be lung cancer, which is the leading cause of cancer deaths in developed countries such as Japan. Numerous observational studies in different ethnic groups have almost consistently shown that BMI is inversely associated with the risk of lung cancer. 3 , 4 , 5 , 6 , 7 , 8 , 9 , 10

Due to the nature of observational studies, it remains controversial whether the inverse association between BMI and lung cancer is causal or merely spurious. Some previous studies reported that this association could be due in part to methodological weaknesses, as it disappeared after restricting analyses to never‐smokers and/or properly adjusting for cigarette smoking (the dominant risk factor for lung cancer). 8 , 10 However, increasing evidence supports the existence of an inverse association, with a 2019 pooled analysis of 12 cohort studies from the United States, Europe, and Asia demonstrating that high BMI was associated with a decreased risk of lung cancer in both smokers and never smokers after excluding cases diagnosed during the first 5 years of follow‐up. 3 As concluded by the authors of this pooled analysis, “the inverse BMI–lung cancer association is not entirely due to smoking and reverse causation.”

Fewer studies have explored the association between BMI and lung cancer in Asian people, who are known to have a lower prevalence of obesity than Caucasians, as well as different body composition. In Japan, three cohort studies 7 , 11 , 12 and two case–control studies 13 , 14 examined the relationship between BMI and lung cancer incidence or mortality, but their findings were inconclusive. Of note, the Japan Public Health Center–based Prospective Study followed 92,098 men and women for an average of 19.1 years and found that after adjustment for smoking and other confounders, lower BMI (<19) at baseline was associated with a 48% increased risk of lung cancer in men. 7 The individual studies mentioned above, however, did not undertake stratified analyses by smoking status, perhaps because of the small number of lung cancer cases in never smokers. This limitation points to the need to conduct a pooled analysis of available cohort study data to further address confounding by smoking.

To better understand the associations between BMI and lung cancer risk, we pooled data from 10 prospective cohort studies with a total of 444,143 participants (including 6454 lung cancer cases). This constituted the largest sample size ever established in the Japanese population. Additionally, we aimed to stratify the associations by sex, smoking status, and histological type.

2. MATERIALS AND METHODS

2.1. Study population

This analysis is part of an ongoing project aiming to elucidate the associations of major cancers with lifestyle factors in Japanese people by pooling data from ongoing cohort studies. A detailed description of the pooling project was provided elsewhere. 15 Briefly, we pooled data from 10 prospective cohort studies that met the predefined criteria: namely the Japan Public Health Center–based Prospective Study (JPHC‐I and JPHC‐II), 16 the Japan Collaborative Cohort Study (JACC), 17 the Miyagi Cohort Study (MIYAGI), 18 the Three‐Prefecture Cohort Study in Miyagi (3‐Pref MIYAGI), 19 the Three‐Prefecture Cohort Study in Aichi (3‐Pref AICHI), 19 the Three‐Prefecture Cohort Study in Osaka (3‐Pref OSAKA), 19 the Ohsaki Cohort Study (OHSAKI), 20 the Takayama Study (TAKAYAMA), 21 and the Life Span Study (LSS). 22 Each cohort enrolled more than 30,000 participants in the 1980s–1990s and used validated questionnaires to collect baseline information on height, weight, cigarette smoking, alcohol consumption, and other lifestyle factors. Selected characteristics of these cohort studies are shown in Table 1.

TABLE 1.

Characteristics of cohort studies included in the present pooled analysis to determine associations between body mass index (BMI) and lung cancer.

Study Cohort participants Age at baseline survey (years) Baseline survey (year) Cohort size Response rate of the baseline questionnaire (%) Outcome ascertainment For the present pooled analysis
Age (years) Date of last follow‐up Mean duration of follow‐up (years) Mean of BMI (SD) Baseline cohort size Lung cancer cases
Men Women Men Women
JPHC‐I Japanese residents of five public health center areas in Japan 40–59 1990 61,595 82 Cancer registry and death certificate 40–59 Dec 31, 2013 20.8 23.6 (3.0) 20,155 21,673 584 219
JPHC‐II Japanese residents of six public health center areas in Japan 40–69 1993‐1994 78,825 80 Cancer registry and death certificate 40–69 Dec 31, 2013 (Suita, Osaka is only until Dec 31, 2012) 17.2 23.4 (3.0) 28,874 31,966 894 330
JACC Residents from 45 areas throughout Japan 40–79 1988‐1990 110,792 83 Cancer registry (selected areas: 24) and death certificate 40–79 2009 13.2 22.8 (3.0) 46,395 64,190 646 226
MIYAGI Residents of 14 municipalities in Miyagi Prefecture, Japan 40–64 1990 47,605 92 Cancer registry and death certificate 40–64 Dec 31, 2014 21.5 23.7 (2.9) 21,094 22,657 800 303
OHSAKI Beneficiaries of National Health Insurance among residents of 14 municipalities in Miyagi Prefecture, Japan 40–79 1994 54,996 95 Cancer registry and death certificate 40–79 Mar 31, 2008 10.8 23.6 (3.1) 21,514 23,209 583 179
TAKAYAMA Residents of Takayama, Gifu Prefecture, Japan ≥35 1992 31,552 92 Cancer registry and death certificate 35–101 Mar 31, 2008 13.7 22.2 (2.8) 13,409 15,573 266 104
3‐Pref MIYAGI Residents of three municipalities in Miyagi Prefecture, Japan 40+ 1984 31,345 94 Cancer registry and death certificate 40+ Dec 31, 1992 7.7 23.2 (3.1) 13,010 15,944 154 56
3‐Pref AICHI Residents of two municipalities in Aichi Prefecture, Japan 40+ 1985 33,529 90 Cancer registry and death certificate 40–103 Dec 31, 2000 11.6 22.1 (2.9) 15,746 17,783 298 115
3‐Pref OSAKA Residents of four municipalities in Osaka Prefecture, Japan 40+ 1983 35,755 82 Cancer registry and death certificate 40–97 Jan 31, 2000 12.3 22.4 (3.0) 15,919 17,973 333 108
LSS Atomic bomb survivors in Hiroshima and Nagasaki, Japan 46–104 1991 20,147 100 Cancer registry and death certificate 46–100 Dec 31, 2003 10.9 22.4 (3.2) 6541 10,518 169 87
Total 202,657 241,486 4727 1727
Open in a new tab

Abbreviations: 3‐Pref AICHI, Three‐Prefecture Study – Aichi portion; 3‐Pref MIYAGI, Three‐Prefecture Study – Miyagi portion; 3‐Pref OSAKA, Three‐Prefecture Study – Osaka portion; JACC, Japan Collaborative Cohort Study; JPHC, Japan Public Health Center–based prospective Study; LSS, Life Span Study; MIYAGI, Miyagi Cohort Study; OHSAKI, Ohsaki National Health Insurance Cohort Study; TAKAYAMA, Takayama Study.

2.2. Body mass index assessment

Self‐reported data on weight and height at baseline were collected from all cohort participants. Body mass index, calculated as weight (kg) divided by the square of height (m), was categorized into six groups: <18.5, 18.5–20.9, 21.0–22.9, 23.0–24.9, 25.0–29.9, and ≥30. Based on the WHO classifications, a BMI of <18.5 is defined as underweight, 18.5–22.9 as normal weight, 25.0–29.9 as overweight, and ≥30 as obese. Individuals with an extreme BMI (<14 or >40) were excluded from the analyses. The validity of self‐reported height and weight was examined in some of the cohorts included in this pooled analysis, with correlation coefficients between self‐reported and measured values ranging from 0.85 to 0.97. 23 , 24 , 25 , 26

2.3. Follow‐up and outcome ascertainment

Follow‐up and outcome ascertainment were carried out according to each cohort's protocol. Cancer diagnoses were confirmed mainly through linkage with cancer registries, review of medical records, or a combination of the two. Lung cancer was ascertained by the International Classification of Diseases (162 in ICD‐9 and C34 in ICD‐10), and was further classified into squamous cell carcinoma (8050–8078, 8083–8084), adenocarcinoma (8140, 8211, 8230–8231, 8250–8260, 8323, 8480–8490, 8550–8552, 8570–8574, 8576), and small‐cell carcinoma (8041–8045) according to the histological grouping proposed by the IARC. 27

2.4. Statistical analysis

Our two‐stage analysis followed the same analytical approach as the one used in the previously mentioned study. 15 In the first stage, individuals were excluded from analyses if they reported a history of any cancer at baseline and had missing data on body weight and/or height. Life Span Study participants with an atomic bomb radiation dose ≥100 mGy were also excluded. We calculated study‐specific HRs of lung cancer in relation to various BMI categories for each participating cohort. Individuals with a BMI of 21.0–22.9 served as the reference group. Hazard ratios and 95% CIs were estimated from Cox proportional hazards regression models. The BMI–lung cancer associations were examined in overall as well as sex‐specific analyses. Assuming a log‐linear dose–response relationship, BMI was also modeled as a continuous variable; HRs were estimated for each 5 kg/m2 increase in BMI. All multivariate models were adjusted for age at baseline, PY of cigarette smoking (men: never, former and current [PY 0< and ≤20], former and current [PY 20< and ≤30], former and current [PY 30< and ≤40], former and current [PY >40]; women: never, former and current [PY 0< and ≤20], former and current [PY 20< and ≤30], former and current [PY >30]), and alcohol consumption (never, occasional, current <23 g/day of ethanol, and current ≥23 g/day of ethanol), with model 3 further adjusting for physical activity if available. For former smokers, we further adjusted for years since quitting smoking (<5, 5 to <10, 10 to <15, and ≥15 years).

In the second stage, we undertook a random‐effects meta‐analysis combining results (β and SE) from each cohort. Heterogeneity across studies was evaluated using Cochran's Q and I 2 statistics. The effect size values were log‐transformed prior to the pooled analysis in order to resolve the asymmetry of the confidence intervals, and then exponentially transformed to obtain the combined values.

In subgroup analyses, we evaluated BMI–lung cancer associations by smoking status (current, former, and never) and histological type (adenocarcinoma, squamous cell carcinoma, and small‐cell carcinoma). The three histological subtypes account for approximately 90% of lung cancers in Japanese patients. 28 , 29 The interaction between smoking and low BMI (<21) in influencing lung cancer risk was evaluated using likelihood tests with the addition of an interaction term. To address reverse causation, we undertook sensitivity analyses that excluded lung cancer cases diagnosed during the first 5 years of follow‐up.

Statistical tests were two‐sided, and a p value <0.05 was considered statistically significant. Analyses were carried out using SAS (version 9.4; SAS Institute) and Stata 17 (StataCorp).

3. RESULTS

Table 1 presents the baseline characteristics of each cohort study included in the current pooled analysis. The mean BMI at baseline ranged from 22.1 to 23.7, and the follow‐up duration ranged from 7.7 to 21.5 years. During 5,730,013 person‐years of follow‐up, 6454 incident lung cancer cases (4727 men and 1727 women) were identified.

Overall, BMI was inversely associated with lung cancer risk; while underweight (BMI <18.5) was associated with a higher risk of lung cancer (HR 1.35; 95% CI, 1.16–1.57), overweight (BMI 25.0–29.9) and obesity (BMI ≥30) were associated with a lower risk, with HRs of 0.77 (95% CI, 0.71–0.84) and 0.69 (95% CI, 0.45–1.07), respectively (Table 2). This inverse association was consistently observed across cohorts included in the pooled analysis, with no significant heterogeneity in risk in the lowest BMI category (BMI <18.5) compared with the reference category (BMI 21.0–22.9). In multivariable‐adjusted models, every 5 kg/m2 increase in BMI was associated with a 21% decrease in the risk of lung cancer (HR 0.79; 95% CI, 0.75–0.83; p < 0.0001). Stratification by sex indicated that the inverse BMI–lung cancer association was similar between men and women, except that the increased risk in underweight (BMI <18.5) women was not statistically significant.

TABLE 2.

Associations between body mass index (BMI) and lung cancer by sex in the Japanese population (N = 444,143).

Total <18.5 18.5 to <21 21 to <23 23 to <25 25 to <30 ≥30 Heterogeneity for lowest category Trend (per 5 kg/m2) Heterogeneity for trend
HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) I 2 (%) Cochran's Q test HR (95% CI) p I 2 (%) Cochran's Q test
Both sexes
Participants, n 392,533 21,407 79,331 104,316 94,931 84,448 8100 – – – – – –
Person‐years 5,733,351 254,390 1,104,802 1,527,754 1,430,791 1,292,930 122,685 – – – – – –
Cases, n 6454 478 1464 1816 1469 1139 88 – – – – – –
HR1 (model 1) a 1.43 (1.29–1.59) 1.12 (1.04–1.21) 1 (Ref.) 0.85 (0.80–0.91) 0.76 (0.71–0.83) 0.73 (0.54–1.01) 0.0 p = 0.38 0.75 (0.71–0.79) <0.0001 30.5 p = 0.16
HR2 (model 2) b 1.37 (1.23–1.52) 1.11 (1.02–1.20) 1 (Ref.) 0.87 (0.81–0.93) 0.77 (0.71–0.83) 0.72 (0.52–1.00) 0.0 p = 0.31 0.77 (0.73–0.81) <0.0001 32.2 p = 0.15
HR3 (model 3) c 1.35 (1.16–1.57) 1.06 (0.98–1.14) 1 (Ref.) 0.87 (0.81–0.93) 0.77 (0.71–0.84) 0.69 (0.45–1.07) 31.0 p = 0.13 0.79 (0.75–0.83) <0.0001 26.6 p = 0.23
Men
Participants, n 180,495 8299 35,616 48,928 46,885 38,021 2746 – – – – – –
Person‐years 2589,596 92,967 482,489 700,211 696,356 576,191 41,382 – – – – – –
Cases, n 4727 336 1144 1371 1065 760 51 – – – – – –
HR1 (model 1) 1.46 (1.28–1.66) 1.16 (1.06–1.26) 1 (Ref.) 0.81 (0.75–0.89) 0.73 (0.66–0.79) 0.83 (0.61–1.14) 3.0 p = 0.26 0.71 (0.66–0.76) <0.0001 36.1 p = 0.11
HR2 (model 2) 1.40 (1.23–1.60) 1.14 (1.05–1.25) 1 (Ref.) 0.83 (0.77–0.91) 0.74 (0.67–0.81) 0.84 (0.61–1.15) 7.6 p = 0.18 0.73 (0.68–0.78) <0.0001 34.1 p = 0.11
HR3 (model 3) 1.43 (1.19–1.71) 1.10 (1.01–1.20) 1 (Ref.) 0.83 (0.75–0.91) 0.74 (0.67–0.81) 0.84 (0.57–1.25) 34.4 p = 0.11 0.74 (0.68–0.80) <0.0001 45.8 p = 0.07
Women
Participants, n 212,038 13,108 43,715 55,388 48,046 46,427 5354 – – – – – –
Person‐years 3,143,755 161,423 622,312 827,542 734,435 716,739 81,303 – – – – – –
Cases, n 1727 142 320 445 404 379 37 – – – – – –
HR1 (model 1) 1.44 (1.17–1.76) 0.98 (0.85–1.13) 1 (Ref.) 0.99 (0.86–1.13) 0.93 (0.77–1.13) 0.83 (0.55–1.24) 0.0 p = 0.32 0.86 (0.79–0.95) 0.003 32.1 p = 0.08
HR2 (model 2) 1.34 (1.10–1.65) 0.96 (0.83–1.11) 1 (Ref.) 1.00 (0.87–1.14) 0.92 (0.76–1.11) 0.78 (0.52–1.18) 1.0 p = 0.24 0.87 (0.79–0.96) 0.0072 39.1 p = 0.05
HR3 (model 3) 1.20 (0.95–1.53) 0.91 (0.78–1.07) 1 (Ref.) 0.99 (0.86–1.15) 0.96 (0.76–1.22) 0.71 (0.43–1.20) 0.0 p = 0.40 0.92 (0.84–1.01) 0.0766 23.2 p = 0.23
Open in a new tab

Note: The interaction between smoking (never vs. ever) and BMI (<21 vs. ≥21) was not statistically significant (p = 0.07 for men and women combined, p = 0.09 for men, and p = 0.12 for women).

Abbreviations: CI, confidence interval; HR, hazard ratio; I 2, inconsistency index; p, probability; Ref. reference.

a

Adjustment for age (year, continuous) and area (for Japan Public Health Center–based prospective Study [JPHC]‐I, JPHC‐II, Japan Collaborative Cohort Study, and Life Span Study).

b

Further adjustment for smoking (men: never, former and current [pack years (PY) 0< and ≤20], former and current [PY 20< and ≤30], former and current [PY 30< and ≤40], former and current [PY >40]; women: never, former and current [PY 0< and ≤20], former and current [PY 20< and ≤30], former and current [PY >30]), drinking (noncurrent drinkers [never‐ or ex‐drinker], occasional drinkers [less than once per week], regular drinkers [<23 g/day], regular drinkers [≥23 g/day]) and history of diabetes (no, yes) to model 1.

c

Further adjustment for exercise to model 2.

Table 3 shows the sex‐specific associations between BMI and lung cancer risk by smoking status. Overall, higher BMI was associated with a lower risk of lung cancer regardless of smoking status, with never, former, and current smokers showing 10%, 22%, and 16% decreases in risk, respectively, per 5 kg/m2 increase in BMI. Notably, overweight (BMI 25.0–29.9) was significantly associated with decreased lung cancer risk in both never‐smokers and ever‐smokers (former and current smokers), with HRs ranging from 0.72 to 0.89. Decreased risk was also observed in obese people regardless of smoking status, but the associations were not statistically significant. Further sex‐stratified analyses by smoking status showed significant, inverse associations between BMI and lung cancer risk in male current and former smokers, but not in male never smokers. In women, no significant, inverse associations were noted in either never smokers or ever smokers (former and current smokers). In addition, we found no statistically significant interaction between smoking and low BMI (p = 0.07 for men and women combined, p = 0.09 for men, and p = 0.13 for women).

TABLE 3.

Associations between body mass index and lung cancer by smoking status in the Japanese population (N = 444,143).

Total <18.5 18.5 to <21 21 to <23 23 to <25 25 to <30 ≥30 Heterogeneity for lowest category Trend (per 5 kg/m2) Heterogeneity for trend
HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) I 2 (%) Cochran's Q test HR (95% CI) p I 2 (%) Cochran's Q test
Both sexes
Never smokers
Participants, n 200,921 10,752 39,546 52,800 48,037 45,080 4706 – – – – – –
Person‐years 3,058,481 136,156 575,205 806,960 751,826 714,960 73,374 – – – – – –
Cases, n 1457 74 235 402 391 331 24 – – – – – –
HR1 (model 1) a – 1.10 (0.84–1.43) 0.87 (0.74–1.03) 1 (Ref.) 0.96 (0.84–1.10) 0.85 (0.75–0.96) 0.70 (0.49–1.00) 18.7 p = 0.17 0.91 (0.83–0.98) 0.019 12.4 p = 0.34
HR2 (model 2) b – 1.09 (0.84–1.40) 0.87 (0.74–1.03) 1 (Ref.) 0.96 (0.84–1.10) 0.85 (0.75–0.96) 0.69 (0.49–0.99) 15.0 p = 0.19 0.90 (0.83–0.98) 0.015 10.3 p = 0.37
HR3 (model 3) c – 1.14 (0.88–1.47) 0.85 (0.71–1.03) 1 (Ref.) 0.95 (0.82–1.10) 0.84 (0.74–0.95) 0.70 (0.49–1.00) 8.4 p = 0.45 0.90 (0.83–0.98) 0.014 11.0 p = 0.54
Former smokers
Participants, n 48,543 2512 9013 12,788 12,773 10,652 805 – – – – – –
Person‐years 678,471 26,351 119,463 178,914 184,698 156,691 12,353 – – – – – –
Cases, n 765 63 186 208 168 131 9 – – – – – –
HR1 (model 1) – 1.56 (1.15–2.10) 1.26 (1.02–1.56) 1 (Ref.) 0.81 (0.63–1.04) 0.75 (0.60–0.94) 1.16 (0.59–2.28) 0.0 p = 0.71 0.75 (0.66–0.86) <0.0001 0.0 p = 0.69
HR2 (model 2) – 1.42 (1.05–1.91) 1.20 (0.91–1.57) 1 (Ref.) 0.83 (0.69–1.01) 0.75 (0.61–0.93) 1.09 (0.60–1.96) 0.0 p = 0.47 0.78 (0.68–0.88) 0.0001 8.2 p = 0.39
HR3 (model 3) – 1.38 (0.90–2.11) 1.05 (0.77–1.44) 1 (Ref.) 0.83 (0.67–1.03) 0.72 (0.57–0.90) 0.91 (0.48–1.73) 39.9 p = 0.13 0.78 (0.67–0.91) 0.0012 20.4 p = 0.26
Current smokers
Participants, n 108,972 6250 24,366 30,113 26,041 20,508 1694 – – – – – –
Person‐years 1,552,638 72,604 331,253 429,282 385,705 308,687 25,108 – – – – – –
Cases, n 3830 318 954 1087 818 606 47 – – – – – –
HR1 (model 1) – 1.42 (1.23–1.65) 1.11 (1.01–1.21) 1 (Ref.) 0.89 (0.81–0.98) 0.88 (0.79–0.97) 1.00 (0.72–1.40) 17.7 p = 0.29 0.81 (0.76–0.87) <0.0001 27.9 p = 0.24
HR2 (model 2) – 1.40 (1.21–1.63) 1.11 (1.01–1.22) 1 (Ref.) 0.89 (0.81–0.98) 0.87 (0.79–0.97) 1.00 (0.72–1.40) 17.9 p = 0.26 0.82 (0.76–0.88) <0.0001 29.5 p = 0.23
HR3 (model 3) – 1.37 (1.12–1.69) 1.08 (0.98–1.19) 1 (Ref.) 0.89 (0.84–0.94) 0.89 (0.80–0.99) 0.98 (0.66–1.44) 39.9 p = 0.11 0.84 (0.78–0.91) <0.0001 24.8 p = 0.25
Men
Never smokers
Participants, n 34,481 1157 5471 8614 9683 8846 710 – – – – – –
Person‐years 491,202 13,966 78,120 131,337 151,293 108,787 7699 – – – – – –
Cases, n 265 8 33 72 94 53 5 – – – – – –
HR1 (model 1) – 1.06 (0.46–2.45) 0.75 (0.49–1.15) 1 (Ref.) 1.16 (0.82–1.64) 0.75 (0.52–1.09) 1.21 (0.47–3.10) 0.0 p = 0.89 0.98 (0.79–1.22) 0.87 0.0 p = 0.45
HR2 (model 2) – 0.89 (0.38–2.11) 0.73 (0.48–1.13) 1 (Ref.) 1.17 (0.84–1.65) 0.77 (0.53–1.11) 1.20 (0.47–3.09) 0.0 p = 0.93 1.02 (0.82–1.26) 0.88 0.0 p = 0.47
HR3 (model 3) – 0.66 (0.22–2.01) 0.71 (0.44–1.12) 1 (Ref.) 1.09 (0.77–1.55) 0.72 (0.49–1.06) 1.18 (0.46–3.04) 0.0 p = 0.99 0.99 (0.79–1.24) 0.92 0.0 p = 0.67
Former smokers
Participants, n 41,326 1884 7349 11,025 11,326 9167 575 – – – – – –
Person‐years 575,961 19,218 96,515 153,428 163,411 134,585 8805 – – – – – –
Cases, n 668 51 159 192 150 109 7 – – – – – –
HR1 (model 1) – 1.47 (1.00–2.17) 1.19 (0.93–1.52) 1 (Ref.) 0.78 (0.61–0.99) 0.70 (0.55–0.89) 1.32 (0.62–2.83) 21.5 p = 0.30 0.74 (0.64–0.85) <0.0001 0.0 p = 0.63
HR2 (model 2) – 1.42 (1.01–1.98) 1.18 (0.89–1.50) 1 (Ref.) 0.82 (0.67–1.00) 0.77 (0.61–0.96) 1.12 (0.59–2.13) 10.2 p = 0.4 0.78 (0.68–0.90) 0.0005 10.4 p = 0.33
HR3 (model 3) – 1.55 (1.05–2.29) 1.05 (0.76–1.46) 1 (Ref.) 0.82 (0.66–1.02) 0.73 (0.56–0.95) 0.88 (0.43–1.80) 15.4 p = 0.3 0.77 (0.65–0.91) 0.0016 21.5 p = 0.25
Current smokers
Participants, n 95,222 4861 21,320 26,953 23,297 17,546 1245 – – – – – –
Person‐years 1,374,002 56,775 294,761 389,834 348,290 265,890 18,452 – – – – – –
Cases, n 3529 263 905 1029 759 537 36 – – – – – –
HR1 (model 1) – 1.36 (1.17–1.57) 1.11 (1.00–1.22) 1 (Ref.) 0.89 (0.81–0.97) 0.87 (0.78–0.97) 1.06 (0.69–1.63) 7.6 p = 0.38 0.82 (0.77–0.88) <0.0001 7.8 p = 0.4
HR2 (model 2) – 1.34 (1.15–1.55) 1.11 (1.00–1.23) 1 (Ref.) 0.89 (0.81–0.97) 0.86 (0.78–0.96) 1.06 (0.70–1.63) 8.4 p = 0.32 0.82 (0.77–0.88) <0.0001 10.5 p = 0.37
HR3 (model 3) – 1.33 (1.08–1.64) 1.07 (0.97–1.18) 1 (Ref.) 0.88 (0.80–0.97) 0.87 (0.78–0.98) 1.11 (0.66–1.85) 33.9 p = 0.12 0.84 (0.79–0.90) <0.0001 3.5 p = 0.32
Women
Never smokers
Participants, n 166,440 9595 34,075 44,186 38,354 36,234 3996 – – – – – –
Person‐years 2,530,699 122,190 497,086 675,623 600,533 572,880 62,387 – – – – – –
Cases, n 1192 66 202 330 297 278 19 – – – – – –
HR1 (model 1) – 1.04 (0.79–1.37) 0.86 (0.71–1.04) 1 (Ref.) 0.97 (0.83–1.14) 0.94 (0.78–1.14) 0.70 (0.43–1.14) 0.0 p = 0.19 0.93 (0.85–1.02) 0.1400 0.0 p = 0.14
HR2 (model 2) – 1.04 (0.79–1.37) 0.86 (0.71–1.04) 1 (Ref.) 0.97 (0.83–1.13) 0.94 (0.77–1.14) 0.69 (0.43–1.13) 0.0 p = 0.20 0.93 (0.85–1.02) 0.1100 0.0 p = 0.13
HR3 (model 3) – 1.09 (0.80–1.48) 0.85 (0.68–1.05) 1 (Ref.) 0.98 (0.82–1.18) 0.94 (0.75–1.20) 0.70 (0.43–1.13) 0.0 p = 0.60 0.91 (0.83–1.01) 0.0700 0.0 p = 0.74
Former smokers
Participants, n 7217 628 1664 1763 1447 1485 230 – – – – – –
Person‐years 102,509 7133 22,948 25,486 21,287 22,106 3549 – – – – – –
Cases, n 97 12 27 16 18 22 2 – – – – – –
HR1 (model 1) – 3.62 (0.56–23.52) 1.32 (0.59–2.95) 1 (Ref.) 1.21 (0.53–2.80) 1.22 (0.50–2.96) 1.90 (0.34–10.68) 56.2 p = 0.11 0.81 (0.57–1.15) 0.2300 16.4 p = 0.37
HR2 (model 2) – 2.22 (0.07–76.10) 0.55 (0.22–1.37) 1 (Ref.) 1.20 (0.53–2.76) 0.66 (0.25–1.73) 1.19 (0.24–5.94) 77.9 p = 0.03 0.98 (0.67–1.42) 0.9000 14.2 p = 0.25
HR3 (model 3) – 2.37 (0.05–104.50) 0.58 (0.22–1.50) 1 (Ref.) 1.20 (0.44–3.26) 0.67 (0.21–2.15) 1.37 (0.20–9.16) 80.3 p = 0.02 1.13 (0.79–1.60) 0.5100 0.0 p = 0.36
Current smokers
Participants, n 15,695 1605 3956 3918 2898 2851 467 – – – – – –
Person‐years 215,053 19,044 53,128 54,131 40,691 41,167 6893 – – – – – –
Cases, n 270 54 61 53 42 51 9 – – – – – –
HR1 (model 1) – 2.17 (1.45–3.26) 1.18 (0.81–1.72) 1 (Ref.) 1.06 (0.70–1.62) 1.10 (0.73‐1.64) 1.62 (0.72–3.63) 0.0 p = 0.95 0.82 (0.66–1.02) 0.0700 33.4 p = 0.17
HR2 (model 2) – 2.16 (1.44–3.25) 1.19 (0.81–1.73) 1 (Ref.) 1.07 (0.70–1.62) 1.08 (0.72‐1.61) 1.56 (0.69–3.52) 0.0 p = 0.93 0.82 (0.65–1.02) 0.0800 35.5 p = 0.14
HR3 (model 3) – 1.98 (1.20–3.28) 1.27 (0.82–1.96) 1 (Ref.) 1.00 (0.61–1.63) 1.17 (0.74‐1.84) 1.18 (0.46–3.05) 0.0 p = 0.83 0.86 (0.65–1.14) 0.3000 45.4 p = 0.09
Open in a new tab

Abbreviations: CI, confidence interval; HR, hazard ratio; I 2, inconsistency index; p, probability; Ref., reference.

a

Adjustment for sex, age (year, continuous) and area (for Japan Public Health Center–based prospective Study [JPHC]‐I, JPHC‐II, Japan Collaborative Cohort Study, and Life Span Study).

b

Further adjustment for drinking (noncurrent drinkers [never‐ or ex‐drinker], occasional drinkers [less than once per week], regular drinkers [<23 g/day], regular drinkers [≥23 g/day]), history of diabetes (no, yes) and years since quitting smoking (<5, 5 to <10, 10 to <15, and ≥15 years) to model 1.

c

Further adjustment for exercise to model 2.

In analyses stratified by histological subtype, an inverse association between BMI and lung cancer risk was evident for adenocarcinoma and squamous cell carcinoma (Table 4), but it was attenuated for small‐cell carcinoma. In further stratification by sex, the associations seemed to vary between men and women. Among the three major subtypes, underweight (BMI <18.5) was significantly associated with an increased risk of adenocarcinoma only among men (HR 1.73; 95% CI, 1.24–2.43), whereas it was significantly associated with an increased risk of squamous cell carcinoma only among women (HR 3.33; 95% CI, 1.60–6.91).

TABLE 4.

Associations between body mass index and lung cancer by histologic type in the Japanese population (N = 444,143).

Total <18.5 18.5 to <21 21 to <23 23 to <25 25 to <30 ≥30 Heterogeneity for lowest category Trend (per 5 kg/m2) Heterogeneity for trend
HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) I 2 (%) Cochran's Q test HR (95% CI) p I 2 (%) Cochran's Q test
Overall
Adenocarcinoma
Participants, n 363,579 19,921 73,821 96,792 87,954 77,709 7382 – – – – – –
Person‐years 5,511,723 244,219 1,063,848 1,470,446 1,376,351 1,239,807 117,051 – – – – – –
Cases, n 2120 112 396 601 534 445 32 – – – – – –
HR1 (model 1) a – 1.34 (1.08–1.66) 1.03 (0.87–1.21) 1 (Ref.) 0.99 (0.84–1.18) 0.77 (0.67–0.87) 0.72 (0.43–1.21) 0 p = 0.74 0.83 (0.76–0.90) <0.0001 13.95 p = 0.55
HR2 (model 2) b – 1.34 (1.08–1.66) 1.02 (0.87–1.20) 1 (Ref.) 0.95 (0.84–1.07) 0.82 (0.70–0.95) 0.76 (0.44–1.31) 0 p = 0.39 0.83 (0.76–0.91) <0.0001 17.89 p = 0.48
HR3 (model 3) c – 1.38 (1.10–1.73) 1.03 (0.87–1.23) 1 (Ref.) 0.93 (0.82–1.05) 0.84 (0.71–0.98) 0.77 (0.45–1.32) 0 p = 0.27 0.83 (0.75–0.91) 0.0001 23.41 p = 0.4
Squamous cell carcinoma
Participants, n 363,579 19,921 73,821 96,792 87,954 77,709 7382 – – – – – –
Person‐years 5511,723 244,219 1063,848 1470,446 1376,351 1239,807 117,051 – – – – – –
Cases, n 1223 75 284 361 284 201 18 – – – – – –
HR1 (model 1) – 1.41 (1.05–1.90) 1.20 (0.95–1.52) 1 (Ref.) 0.79 (0.67–0.93) 0.72 (0.61–0.86) 0.91 (0.56–1.46) 15.01 p = 0.38 0.74 (0.65–0.85) <0.0001 30.08 p = 0.08
HR2 (model 2) – 1.35 (1.02–1.78) 1.19 (0.94–1.51) 1 (Ref.) 0.81 (0.69–0.96) 0.74 (0.62–0.88) 0.88 (0.54–1.42) 3.89 p = 0.51 0.77 (0.69–0.85) <0.0001 0 p = 0.1
HR3 (model 3) – 1.40 (1.04–1.87) 1.09 (0.87–1.35) 1 (Ref.) 0.82 (0.69–0.96) 0.72 (0.60–0.86) 0.81 (0.50–1.33) 5.16 p = 0.38 0.77 (0.69–0.86) <0.0001 0 p = 0.15
Small‐cell carcinoma
Participants, n 363,579 19,921 73,821 96,792 87,954 77,709 7382 – – – – – –
Person‐years 5,511,723 244,219 1,063,848 1,470,446 1,376,351 1,239,807 117,051 – – – – – –
Cases, n 529 31 119 153 109 107 10 – – – – – –
HR1 (model 1) – 1.60 (1.05–2.45) 1.15 (0.89–1.48) 1 (Ref.) 0.74 (0.57–0.95) 0.87 (0.67–1.12) 1.19 (0.62–2.29) 0 p = 0.60 0.78 (0.65–0.94) 0.009 20.14 p = 0.30
HR2 (model 2) – 1.52 (0.99–2.33) 1.11 (0.86–1.44) 1 (Ref.) 0.74 (0.57–0.95) 0.9 (0.69–1.16) 1.16 (0.60–2.24) 0 p = 0.39 0.81 (0.67–0.97) 0.0244 16.72 p = 0.27
HR3 (model 3) – 1.48 (0.93–2.36) 1.10 (0.85–1.44) 1 (Ref.) 0.77 (0.59–1.01) 0.86 (0.66–1.12) 1.17 (0.61–2.26) 0 p = 0.25 0.81 (0.65–1.00) 0.0496 31.89 p = 0.07
Men
Adenocarcinoma
Participants, n 167,485 7734 33,136 45,330 43,466 35,255 2564 – – – – – –
Person‐years 2,491,941 89,176 464,280 673,248 670,166 555,060 40,011 – – – – – –
Cases, n 1185 69 256 333 297 214 16 – – – – – –
HR1 (model 1) – 1.63 (1.25–2.14) 1.20 (1.01–1.42) 1 (Ref.) 0.91 (0.77–1.06) 0.74 (0.62–0.88) 0.94 (0.54–1.64) 0 p = 0.40 0.72 (0.65–0.80) <0.0001 0 p = 0.95
HR2 (model 2) – 1.58 (1.10–2.26) 1.18 (1–1.4) 1 (Ref.) 0.92 (0.78–1.08) 0.75 (0.63–0.90) 0.93 (0.53–1.61) 30.78 p = 0.14 0.73 (0.66–0.82) <0.0001 0 p = 0.79
HR3 (model 3) – 1.73 (1.24–2.43) 1.19 (1–1.42) 1 (Ref.) 0.91 (0.77–1.08) 0.76 (0.63–0.91) 0.93 (0.53–1.62) 18.35 p = 0.24 0.73 (0.65–0.82) <0.0001 0 p = 0.48
Squamous cell carcinoma
Participants, n 167,485 7734 33,136 45,330 43,466 35,255 2564 – – – – – –
Person‐years 2,491,941 89,176 464,280 673,248 670,166 555,060 40,011 – – – – – –
Cases, n 1110 60 258 334 262 180 16 – – – – – –
HR1 (model 1) – 1.25 (0.88–1.76) 1.17 (0.91–1.49) 1 (Ref.) 0.78 (0.66–0.93) 0.79 (0.58–1.09) 1.02 (0.62–1.67) 23.36 p = 0.31 0.77 (0.66–0.91) 0.0020 40.31 p = 0.05
HR2 (model 2) – 1.21 (0.86–1.70) 1.16 (0.90–1.49) 1 (Ref.) 0.81 (0.69–0.96) 0.74 (0.61–0.89) 0.94 (0.57–1.56) 18.38 p = 0.36 0.79 (0.71–0.89) <0.0001 0 p = 0.1
HR3 (model 3) – 1.23 (0.85–1.78) 1.07 (0.85–1.34) 1 (Ref.) 0.81 (0.69–0.97) 0.72 (0.59–0.87) 0.95 (0.57–1.57) 24.46 p = 0.25 0.8 (0.71–0.9) 0.0002 0 p = 0.13
Small‐cell carcinoma
Participants, n 167,485 7734 33,136 45,330 43,466 35,255 2564 – – – – – –
Person‐years 2,491,941 89,176 464,280 673,248 670,166 555,060 40,011 – – – – – –
Cases, n 459 26 101 137 97 91 7 – – – – – –
HR1 (model 1) – 1.59 (1.02–2.49) 1.07 (0.81–1.40) 1 (Ref.) 0.74 (0.56–0.97) 0.88 (0.67–1.15) 1.16 (0.53–2.53) 0 p = 0.88 0.78 (0.65–0.93) 0.0070 6.47 p = 0.67
HR2 (model 2) – 1.43 (0.90–2.26) 1.03 (0.78–1.36) 1 (Ref.) 0.73 (0.56–0.96) 0.88 (0.66–1.16) 1.17 (0.54–2.56) 0 p = 0.86 0.80 (0.67–0.96) 0.0174 4.25 p = 0.47
HR3 (model 3) – 1.45 (0.88–2.38) 1.02 (0.76–1.36) 1 (Ref.) 0.76 (0.58–1.01) 0.90 (0.68–1.19) 1.18 (0.54–2.58) 0 p = 0.58 0.82 (0.68–0.99) 0.0409 4.2 p = 0.36
Women
Adenocarcinoma
Participants, n 196,094 12,187 40,685 51,462 44,488 42,454 4818 – – – – – –
Person‐years 3,019,783 155,044 599,569 797,198 706,185 684,747 77,041 – – – – – –
Cases, n 934 43 140 268 237 231 15 – – – – – –
HR1 (model 1) − 1.14 (0.81–1.60) 0.83 (0.66–1.03) 1 (Ref.) 1.11 (0.84–1.46) 0.98 (0.74–1.29) 0.67 (0.39–1.17) 0 p = 0.91 0.97 (0.84–1.11) 0.6110 29.21 p = 0.16
HR2 (model 2) − 1.15 (0.81–1.63) 0.81 (0.64–1.02) 1 (Ref.) 1.16 (0.85–1.57) 1.02 (0.73–1.41) 0.66 (0.38–1.16) 0 p = 0.68 0.96 (0.84–1.09) 0.5376 22.83 p = 0.21
HR3 (model 3) − 1.10 (0.76–1.60) 0.86 (0.64–1.16) 1 (Ref.) 1.08 (0.78–1.50) 1.14 (0.75–1.74) 0.67 (0.38–1.19) 0 p = 0.65 0.99 (0.85–1.15) 0.8497 38.77 p = 0.11
Squamous cell carcinoma
Participants, n 196,094 12,187 40,685 51,462 44,488 42,454 4818 – – – – – –
Person‐years 3,019,783 155,044 599,569 797,198 706,185 684,747 77,041 – – – – – –
Cases, n 113 15 26 27 22 21 2 – – – – – –
HR1 (model 1) − 3.22 (1.67–6.22) 1.57 (0.92–2.67) 1 (Ref.) 0.83 (0.47–1.47) 0.94 (0.54–1.63) 2.97 (0.63–13.92) 0 p = 0.67 0.60 (0.44–0.83) 0.002 9.73 p = 0.45
HR2 (model 2) − 3.04 (1.51–6.12) 1.11 (0.6–2.04) 1 (Ref.) 0.87 (0.47–1.61) 0.76 (0.41–1.42) 2.74 (0.52–14.47) 0 p = 0.95 0.64 (0.46–0.9) 0.0103 9.87 p = 0.49
HR3 (model 3) − 3.33 (1.60–6.91) 1.10 (0.58–2.1) 1 (Ref.) 0.90 (0.48–1.71) 0.77 (0.40–1.46) 1.72 (0.17–17.20) 0 p = 0.96 0.6 (0.42–0.87) 0.0065 12.26 p = 0.45
Small‐cell carcinoma
Participants, n 196,094 12,187 40,685 51,462 44,488 42,454 4818 – – – – – –
Person‐years 3,019,783 155,044 599,569 797,198 706,185 684,747 77,041 – – – – – –
Cases, n 70 5 18 16 12 16 3 – – – – – –
HR1 (model 1) − 2.42 (0.87–6.75) 2.28 (1.15–4.51) 1 (Ref.) 0.72 (0.33–1.53) 1.19 (0.62–2.30) 2.39 (0.61–9.35) 0 p = 1.00 0.81 (0.51–1.29) 0.3790 35 p = 0.11
HR2 (model 2) − 1.73 (0.42–7.13) 2.18 (0.72–6.60) 1 (Ref.) 0.87 (0.38–1.99) 0.91 (0.4–2.07) 1.95 (0.48–7.99) 0 p = 0.83 0.91 (0.56–1.46) 0.6855 33.53 p = 0.15
HR3 (model 3) − 1.71 (0.33–8.86) 2.25 (0.72–7.00) 1 (Ref.) 0.87 (0.38–1.99) 0.75 (0.33–1.7) 1.95 (0.48–7.98) 0 p = 0.47 0.83 (0.41–1.67) 0.5936 55.86 p = 0.07
Open in a new tab

Abbreviations: CI, confidence interval; HR, hazard ratio; I 2, inconsistency index; p, probability; Ref., reference.

a

Adjustment for age (year, continuous) and area (for JPHC‐I, JPHC‐II, JACC, and LSS).

b

Further adjustment for smoking (men: never, former and current [pack‐years (PY) 0< and ≤20], former and current [PY 20< and ≤30], former and current [PY 30< and ≤40], former and current [PY >40]; women: never, former and current [PY 0< and ≤20], former and current [PY 20< and ≤30], former and current [PY >30]), drinking (noncurrent drinkers [never‐ or ex‐drinker], occasional drinkers [less than once per week], regular drinkers [<23 g/day], regular drinkers [≥23 g/day]) and history of diabetes (no, yes) to model 1.

c

Further adjustment for exercise to model 2.

To address the effect of reverse causation, we repeated the analyses, this time excluding lung cancer cases diagnosed during the first 5 years of follow‐up. Overall, the inverse associations persisted for both men and women; however, further analyses stratified by smoking status suggested that the associations were more apparent in male former and current smokers, with no significant trend in female nonsmokers (Tables S1–S3).

4. DISCUSSION

In this pooled analysis of 10 prospective cohort studies, we observed an overall inverse association between BMI and the risk of lung cancer in Japanese people, with each 5 kg/m2 increase in BMI associated with 21% decreased risk. This estimate was consistent with a number of previous cohort studies, meta‐analyses, and pooled analyses that included ethnically diverse populations. 3 , 4 , 5 , 6 , 7 , 8 , 9 , 10 Notably, our estimate of a 21% decreased risk per 5 kg/m2 increase in BMI was larger than that (11%) reported in the 2018 pooled analysis of 30 prospective cohort studies involving more than 1.6 million individuals from the United States, Europe, and Asia. 3

While obesity has been causally linked with at least 13 cancer sites, its causal relationship with lung cancer remains inconclusive. Although almost all observational studies (including ours) have consistently shown an inverse association, it remains controversial whether this was due to a real cause–effect relation or whether it merely represents a spurious association stemming from confounding, reverse causation, or other sources of bias. Among confounding factors that could distort BMI–lung cancer associations, cigarette smoking is a major concern because current smokers are well known to weigh less and have a substantially higher risk of developing lung cancer than never smokers. Previous studies that sought to control for smoking yielded mixed findings, with some showing the attenuation or even disappearance of inverse associations in analyses restricted to never‐smokers. 8 , 10 The 2018 pooled analysis provided evidence supporting the inverse association in both smokers and never smokers. 3 Our findings showing an overall inverse association in men and women combined were largely consistent with the results of the 2018 pooled analysis. However, further sex‐specific analyses revealed that the increased risk associated with low BMI was apparent only in male current and former smokers, and was attenuated or absent in male never smokers. Among women, low BMI was associated with an increased risk of lung cancer, but this association was not statistically significant. One possible reason for the different results between sexes is the limited statistical power of the sex‐specific analyses. Together, our findings suggest that the overall inverse associations between BMI and lung cancer risk may be partly driven by current and former male smokers. In addition, we examined the interaction between BMI and smoking in influencing lung cancer risk and found no statistically significant interaction between low BMI and ever smoking. Given that previous studies have shown increased levels of 8‐OHdG (endogenous oxidative damage to DNA) in lean smokers and an inverse association between weight loss and 8‐OHdG levels, 30 , 31 leanness may contribute to reduced biological functions against smoking‐induced oxidative DNA damage. Further studies are needed to address biological or statistical interactions between smoking and BMI in modulating lung cancer risk.

Another possible explanation for the observed inverse associations is reverse causation, which refers to the fact that undiagnosed lung carcinoma or other chronic medical conditions precede and cause weight loss. 32 To evaluate the effect of reverse causation on risk estimates, we repeated the analyses but excluded cases diagnosed during the first 5 years of follow‐up. The results remained materially unchanged when compared to the main analyses, thereby implying that the observed inverse associations could not be explained by pre‐existing illness.

Even with careful attempts to address confounding and reverse causation, establishing an inverse causal relationship between BMI and lung cancer risk remains challenging because of the nature of observational studies as well as the lack of mechanistic understanding. The emergence of MR—an approach using genetic variants as instrumental variables to approximate environmental exposure—has offered a solution to circumvent the limitations (confounding and reverse causation) that are inherent in observational studies. 33 Several MR studies revealed that genetically predicted BMI was associated with an increased risk of lung cancer of all types, 34 , 35 , 36 , 37 , 38 a finding that contrasts with the inverse association seen in observational studies. One interpretation is that MR studies evaluated the association of lung cancer with static, genetically determined BMI throughout one's lifetime, while observational studies often examined associations with one single baseline BMI measurement in adulthood. Another reason may be that BMI‐associated genetic variants used in these MR studies were derived from genome‐wide association studies involving populations of European ancestry, making it uncertain whether the results can be generalized to populations of East Asian ancestry. Further MR analyses using genetic variants associated with BMI in the Japanese population are warranted to corroborate or refute BMI–lung cancer associations.

Whether the association of BMI with lung cancer differs by histological subtype has been explored in both observational and MR studies, but the findings are inconclusive. The 2018 pooled analysis found inverse associations for adenoma and squamous cell carcinoma, but identified a positive association for small‐cell lung cancer. 3 In our pooled analysis, similar inverse associations were observed for adenocarcinoma and squamous cell carcinoma; however, no significant associations were noted for small‐cell lung cancer. In addition, the results of MR studies were not entirely consistent, with inverse association documented for lung adenocarcinoma in some studies. 34 , 39 These findings indicate that BMI may exert differential effects on lung cancer histological subtypes in different ethnic groups.

A major strength of our study is that the number of lung cancer cases in the Japanese population was larger than that in any previous study. By pooling data, we were able to use the same BMI categories and covariate definitions, analyzing the associations between BMI and the risk of lung cancer according to sex, smoking status, and histological type. In addition, we were able to address the effect of confounding by adjusting for PY of smoking and restricting the analysis to never‐smokers.

Our study also has limitations. First, BMI is known to be an imperfect measurement of adiposity; it does not distinguish between adipose tissue and lean body mass, nor does it reflect metabolic or endocrine disruptions associated with obesity. 40 However, few studies have explored whether other measures of fatness and body composition, such as waist circumference and waist‐to‐hip ratio, are associated with the risk of lung cancer in Japan. 7 Second, we were only able to analyze a single baseline BMI measurement for each individual; it is possible that changes in weight and height over time may have influenced the observed associations. Third, we acknowledge the lack of statistical power for certain subgroup analyses, such as histological type, because of the small number of cases. The possibility that some of the results of these analyses were due to chance cannot be ruled out. Fourth, despite our best efforts to address confounders, the confounding effect of smoking may have persisted, in particular in lean male smokers, and other unknown confounders might also have distorted the observed associations. Finally, compared with current smokers, the precise risk estimate for underweight never smokers is still challenging because they are thought to be a heterogeneous group; their BMI is thought to be influenced by various factors, including genetics, passive smoking, underlying medical conditions, culture, and socioeconomic status. 41 Further refinement in risk estimates is needed for this group.

In summary, our findings add to the evidence that low BMI is inversely associated with an increased risk of lung cancer. This may be driven by current and former smokers, and warrants further investigation of never‐smokers in additional studies.

AUTHOR CONTRIBUTIONS

Sayo Kawai: Conceptualization; data curation; formal analysis; methodology; writing – original draft; writing – review and editing. Yingsong Lin: Conceptualization; data curation; methodology; project administration; supervision; writing – original draft; writing – review and editing. Hiroshi Tsuge: Formal analysis; methodology; resources; writing – review and editing. Hidemi Ito: Data curation; resources; supervision; writing – review and editing. Keitaro Matsuo: Data curation; investigation; project administration; supervision; validation; writing – review and editing. Keiko Wada: Data curation; formal analysis; investigation; project administration; resources; validation; writing – review and editing. Chisato Nagata: Investigation; project administration; resources; supervision; validation; writing – review and editing. Nobuhiro Narii: Data curation; formal analysis; validation; writing – review and editing. Tetsuhisa Kitamura: Data curation; investigation; project administration; supervision; validation; writing – review and editing. Mai Utada: Data curation; formal analysis; investigation; project administration; resources; validation; writing – review and editing. Ritsu Sakata: Investigation; project administration; resources; validation. Takashi Kimura: Data curation; formal analysis; validation; writing – review and editing. Akiko Tamakoshi: Data curation; investigation; project administration; resources; supervision; writing – review and editing. Yumi Sugawara: Data curation; formal analysis; validation; writing – review and editing. Ichiro Tsuji: Investigation; project administration; resources; supervision; writing – review and editing. Seitaro Suzuki: Data curation; formal analysis; validation; writing – review and editing. Norie Sawada: Investigation; project administration; resources; supervision; writing – review and editing. Shoichiro Tsugane: Investigation; project administration; resources; supervision; writing – review and editing. Tetsuya Mizoue: Conceptualization; methodology; supervision; writing – review and editing. Isao Oze: Investigation; project administration; supervision; validation; writing – review and editing. Sarah Krull Abe: Conceptualization; methodology; supervision; validation; writing – review and editing. Manami Inoue: Conceptualization; funding acquisition; project administration; resources; supervision; validation; writing – review and editing.

CONFLICT OF INTEREST STATEMENT

Inoue Manami, Keitaro Matsuo, Chisato Nagata, and Norie Sawada are Editorial Board Members of Cancer Science. The other authors have no conflict of interest.

ETHICS STATEMENT

Approval of the research protocol by an institutional review board: Approval of institutional review boards was obtained for each participating study included in the pooled analysis.

Informed consent: Informed consent was obtained from participants of each cohort study.

Registry and the registration no. of the study/trial: N/A.

Animal studies: N/A.

Supporting information

Tables S1–S3.

CAS-115-1346-s001.docx (85.7KB, docx)

ACKNOWLEDGMENTS

This work was supported by the National Cancer Center Research and Development Fund (2021‐A‐16, 30‐A‐15, 27‐A‐4, and 24‐A‐3) and by Health and Labour Sciences Research Grants for the Third Term Comprehensive Control Research for Cancer (H21‐3jigan‐ippan‐003, H18‐3jigan‐ippan‐001, and H16‐3jigan‐010). The Radiation Effects Research Foundation (RERF) is funded by the governments of Japan and the United States (RERF Research Protocol A2‐15).

Kawai S, Lin Y, Tsuge H, et al. Body mass index and lung cancer risk: Pooled analysis of 10 prospective cohort studies in Japan. Cancer Sci. 2024;115:1346‐1359. doi: 10.1111/cas.16093

Research group members are listed at the following site (as of July 2021): http://epi.ncc.go.jp/en/can_prev/796/7955.html.

DATA AVAILABILITY STATEMENT

The data underlying this manuscript cannot be shared publicly due to the privacy of study participants. A collaboration with each participating cohort study is required to access the data.

REFERENCES

  • 1. GBD 2015 Obesity Collaborators , Afshin A, Forouzanfar MH, Reitsma MB, et al. Health effects of overweight and obesity in 195 countries over 25 years. N Engl J Med. 2017;377:13‐27. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2. Lauby‐Secretan B, Scoccianti C, Loomis D, Grosse Y, Bianchini F, Straif K. International Agency for Research on Cancer handbook working group. Body fatness and cancer—viewpoint of the IARC working group. N Engl J Med. 2016;375:794‐798. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3. Yu D, Zheng W, Johansson M, et al. Overall and central obesity and risk of lung cancer: a pooled analysis. J Natl Cancer Inst. 2018;110:831‐842. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4. Yang Y, Dong J, Sun K, et al. Obesity and incidence of lung cancer: a meta‐analysis. Int J Cancer. 2013;132:1162‐1169. [DOI] [PubMed] [Google Scholar]
  • 5. Koh WP, Yuan JM, Wang R, Lee HP, Yu MC. Body mass index and smoking‐related lung cancer risk in the Singapore Chinese health study. Br J Cancer. 2010;102:610‐614. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6. Jee SH, Yun JE, Park EJ, et al. Body mass index and cancer risk in Korean men and women. Int J Cancer. 2008;123:1892‐1896. [DOI] [PubMed] [Google Scholar]
  • 7. Abe SK, Narita S, Saito E, et al. Body mass index, height, weight change, and subsequent lung cancer risk: the Japan public health center‐based prospective study. Cancer Epidemiol Biomarkers Prev. 2021;30:1708‐1716. [DOI] [PubMed] [Google Scholar]
  • 8. Dewi NU, Boshuizen HC, Johansson M, et al. Anthropometry and the risk of lung cancer in EPIC. Am J Epidemiol. 2016;184:129‐139. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9. Smith L, Brinton LA, Spitz MR, et al. Body mass index and risk of lung cancer among never, former, and current smokers. J Natl Cancer Inst. 2012;104:778‐789. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10. Wood AM, Jonsson H, Nagel G, et al. The inverse Association of Body Mass Index with lung cancer: exploring residual confounding, metabolic aberrations and within‐person variability in smoking. Cancer Epidemiol Biomarkers Prev. 2021;30:1489‐1497. [DOI] [PubMed] [Google Scholar]
  • 11. Kondo T, Hori Y, Yatsuya H, et al. Lung cancer mortality and body mass index in a Japanese cohort: findings from the Japan collaborative cohort study (JACC study). Cancer Causes Control. 2007;18:229‐234. [DOI] [PubMed] [Google Scholar]
  • 12. Kuriyama S, Tsubono Y, Hozawa A, et al. Obesity and risk of cancer in Japan. Int J Cancer. 2005;113:148‐157. [DOI] [PubMed] [Google Scholar]
  • 13. Kagohashi K, Satoh H, Kurishima K, Ishikawa H, Ohtsuka M. Body mass index and lung cancer risk in never smokers. Radiol Oncol. 2006;40:239‐244. [Google Scholar]
  • 14. Kanashiki M, Sairenchi T, Saito Y, Ishikawa H, Satoh H, Sekizawa K. Body mass index and lung cancer: a case‐control study of subjects participating in a mass‐screening program. Chest. 2005;128:1490‐1496. [DOI] [PubMed] [Google Scholar]
  • 15. Sasazuki S, Inoue M, Tsuji I, et al. Body mass index and mortality from all causes and major causes in Japanese: results of a pooled analysis of 7 large‐scale cohort studies. J Epidemiol. 2011;21:417‐430. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16. Tsugane S, Sawada N. The JPHC study: design and some findings on the typical Japanese diet. Jpn J Clin Oncol. 2014;44:777‐782. [DOI] [PubMed] [Google Scholar]
  • 17. Tamakoshi A, Yoshimura T, Inaba Y, et al. Profile of the JACC study. J Epidemiol. 2005;15:S4‐S8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18. Tsuji I, Nishino Y, Tsubono Y, et al. Follow‐up and mortality profiles in the Miyagi cohort study. J Epidemiol. 2004;14:S2‐S6. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19. Marugame T, Sobue T, Satoh H, et al. Lung cancer death rates by smoking status: comparison of the three‐prefecture cohort study in Japan to the cancer prevention study II in the USA. Cancer Sci. 2005;96:120‐126. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20. Tsuji I, Nishino Y, Ohkubo T, et al. A prospective cohort study on National Health Insurance beneficiaries in Ohsaki, Miyagi prefecture, Japan: study design, profiles of the subjects and medical cost during the first year. J Epidemiol. 1998;8:258‐263. [DOI] [PubMed] [Google Scholar]
  • 21. Shimizu H. The Basic Report on Takayama Study. Department of Public Health, Gifu University School of Medicine; 1996. [Google Scholar]
  • 22. Ozasa K, Grant EJ, Kodama K. Japanese legacy cohorts: the life span study atomic bomb survivor cohort and Survivors' offspring. J Epidemiol. 2018;28:162‐169. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23. Shimizu N, Nagata C, Shimizu H, et al. Height, weight, and alcohol consumption in relation to the risk of colorectal cancer in Japan: a prospective study. Br J Cancer. 2003;887:1038‐1043. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24. Inoue M, Sobue T, Tsugane S. Impact of body mass index on the risk of total cancer incidence and mortality among middle‐aged Japanese: data from a large‐scale population‐based cohort study—the JPHC study. Cancer Causes Control. 2004;15:671‐680. [DOI] [PubMed] [Google Scholar]
  • 25. Kuriyama S, Ohmori K, Miura C, et al. Body mass index and mortality in Japan: the Miyagi cohort study. J Epidemiol. 2004;14(Suppl 1):S33‐S38. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26. Kuriyama S, Tsuji I, Ohkubo T, et al. Medical care expenditure associated with body mass index in Japan: the Ohsaki study. Int J Obes Relat Metab Disord. 2002;26:1069‐1074. [DOI] [PubMed] [Google Scholar]
  • 27. Bray F, Colombet M, Mery L, et al. Cancer Incidence in Five Continents, XI (electronic version). International Agency for Research on Cancer; 2017. Accessed March 2023. Available from: https://ci5.iarc.fr/CI5‐XI/Pages/Chapter4.aspx [Google Scholar]
  • 28. Toyoda Y, Nakayama T, Ioka A, Tsukuma H. Trends in lung cancer incidence by histological type in Osaka. Japan Jpn J Clin Oncol. 2008;38:534‐539. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29. Nguyen PT, Katanoda K, Saito E, Hori M, Nakayama T, Matsuda T. Trends in lung cancer incidence by gender, histological type and stage at diagnosis in Japan, 1993 to 2015: a multiple imputation approach. Int J Cancer. 2022;151:20‐32. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30. Mizoue T, Kasai H, Kubo T, Tokunaga S. Leanness, smoking, and enhanced oxidative DNA damage. Cancer Epidemiol Biomarkers Prev. 2006;15:582‐585. [DOI] [PubMed] [Google Scholar]
  • 31. Mizoue T, Tokunaga S, Kasai H, Kawai K, Sato M, Kubo T. Body mass index and oxidative DNA damage: a longitudinal study. Cancer Sci. 2007;98:1254‐1258. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32. Flegal KM, Graubard BI, Williamson DF, Cooper RS. Reverse causation and illness‐related weight loss in observational studies of body weight and mortality. Am J Epidemiol. 2011;173:1‐9. [DOI] [PubMed] [Google Scholar]
  • 33. Holmes MV, Ala‐Korpela M, Smith GD. Mendelian randomization in cardiometabolic disease: challenges in evaluating causality. Nat Rev Cardiol. 2017;14:577‐590. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34. Fang Z, Song M, Lee DH, Giovannucci EL. The role of Mendelian randomization studies in deciphering the effect of obesity on cancer. J Natl Cancer Inst. 2022;114:361‐371. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35. Zhou W, Liu G, Hung RJ, et al. Causal relationships between body mass index, smoking and lung cancer: univariable and multivariable Mendelian randomization. Int J Cancer. 2021;148:1077‐1086. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36. Jiang L, Sun YQ, Brumpton BM, Langhammer A, Chen Y, Mai XM. Body mass index and incidence of lung cancer in the HUNT study: using observational and Mendelian randomization approaches. BMC Cancer. 2022;22:1152. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37. Mariosa D, Smith‐Byrne K, Richardson TG, et al. Body size at different ages and risk of six cancers: a Mendelian randomization and prospective cohort study. J Natl Cancer Inst. 2022;114:1296‐1300. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38. Vithayathil M, Carter P, Kar S, Mason AM, Burgess S, Larsson SC. Body size and composition and risk of site‐specific cancers in the UK biobank and large international consortia: a mendelian randomisation study. PLoS Med. 2021;18:e1003706. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39. Carreras‐Torres R, Johansson M, Haycock PC, 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]
  • 40. Flegal KM, Shepherd JA, Looker AC, et al. Comparisons of percentage body fat, body mass index, waist circumference, and waist‐stature ratio in adults. Am J Clin Nutr. 2009;89:500‐508. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41. Samet JM, Avila‐Tang E, Boffetta P, et al. Lung cancer in never smokers: clinical epidemiology and environmental risk factors. Clin Cancer Res. 2009;15:5626‐5645. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

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

Supplementary Materials

Tables S1–S3.

CAS-115-1346-s001.docx (85.7KB, docx)

Data Availability Statement

The data underlying this manuscript cannot be shared publicly due to the privacy of study participants. A collaboration with each participating cohort study is required to access the data.

Articles from Cancer Science are provided here courtesy of Wiley

RESOURCES