Disparities in Cancer Incidence by Sex and Associated Lifestyle Factors and Health Conditions

Zhuo-Ying Li; Yu-Ting Tan; Qiu-Ming Shen; Dan-Ni Yang; Hong-Lan Li; Yong-Bing Xiang

doi:10.1001/jamaoncol.2026.3747

. 2026 Sep 24:e263747. Online ahead of print. doi: 10.1001/jamaoncol.2026.3747

Disparities in Cancer Incidence by Sex and Associated Lifestyle Factors and Health Conditions

Zhuo-Ying Li ^1,², Yu-Ting Tan ^1,², Qiu-Ming Shen ^1,², Dan-Ni Yang ^1,^2,³, Hong-Lan Li ^1,², Yong-Bing Xiang ^1,^2,^3,^✉

¹State Key Laboratory of Systems Medicine for Cancer, Shanghai Cancer Institute, Renji Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China

²Department of Epidemiology, Shanghai Cancer Institute, Shanghai, China

³School of Public Health, Shanghai Jiao Tong University School of Medicine, Shanghai, China

Accepted for Publication: August 4, 2026.

Published Online: September 24, 2026. doi:10.1001/jamaoncol.2026.3747

Open Access: This is an open access article distributed under the terms of the CC-BY-NC-ND License, which does not permit alteration or commercial use, including those for text and data mining, AI training, and similar technologies. © 2026 Li ZY et al. JAMA Oncology.

^✉

Corresponding Author: Yong-Bing Xiang, MD, MSc, State Key Laboratory of Systems Medicine for Cancer, Shanghai Cancer Institute, Renji Hospital, Shanghai Jiao Tong University School of Medicine, No. 25, Ln 2200, Xie Tu Rd, Shanghai 200032, PR China (ybxiang@shsci.org).

Author Contributions: Drs Zhuo-Ying Li and Xiang had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Z.-Y. Li, Xiang.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Z.-Y. Li, Xiang.

Critical review of the manuscript for important intellectual content: All authors.

Statistical analysis: Z.-Y. Li, Shen, Xiang.

Obtained funding: Z.-Y. Li, Xiang.

Administrative, technical, or material support: Tan, H. Li, Xiang.

Supervision: Xiang.

Conflict of Interest Disclosures: None reported.

Funding/Support: This work was supported by the National Natural Science Foundation of China (82504486), the Shanghai Municipal Bureau of Disease Control and Prevention (2025GKQ04), the National Key Project of Research and Development Program of China (2021YFC2500404), and the US National Institutes of Health (subcontracts of UM1 CA173640, CA182910).

Role of the Funder/Sponsor: The funder had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Data Sharing Statement: See Supplement 2.

^✉

Corresponding author.

PMCID: PMC13613232 PMID: 42783354

Key Points

Question

Given that cancer incidence is higher among males than among females globally, to what extent do modifiable lifestyle behaviors and health conditions contribute to this disparity by sex across diverse populations?

Findings

In this multicohort study of 589 766 adults in China and the UK, 11 measured factors jointly accounted for 84% of the male-female gap in non−sex-specific cancer incidence (183 cases per 100 000 person-years) in the Chinese population but only 26% (52 cases per 100 000 person-years) in the UK population.

Meaning

These findings suggest that the contribution of modifiable lifestyle behaviors and health conditions to the male-female cancer gap varies across populations, underscoring the need for population-specific prevention.

Abstract

Importance

Cancer risk is higher in males than females globally. Whether this sex disparity reflects a biological inevitability or is potentially modifiable by lifestyle behaviors and health conditions across diverse populations remains unclear.

Objective

To determine the contribution of lifestyle factors and health conditions to sex disparities in cancer incidence across populations.

Design, Setting, and Participants

This cohort study conducted an individual-level analysis of 3 population-based prospective cohorts in China (Shanghai Men’s Health Study and Shanghai Women’s Health Study) and the UK (UK Biobank) of adults aged 40 to 74 years at baseline with no prior cancer. Baseline data were collected from 1997 to 2010, with follow-up through 2021. Data were analyzed from September 2025 to June 2026.

Exposures

Eleven risk factors covering behaviors (smoking, heavy alcohol drinking, physical inactivity), diet (low intakes of vegetable and fruit, high intakes of red meat and processed meat), and health conditions (cholelithiasis, chronic liver diseases, diabetes, obesity).

Main Outcomes and Measures

Incidence of 34 non−sex-specific cancer end points. Attributable rate differences (ARDs, cases per 100 000 person-years) and attributable proportion (AP, %) quantified the number and percentage of the male-female incidence rate gap attributable to the exposures.

Results

Among a total of 589 766 participants (mean [SD] age, 56.1 [8.5] years; 315 813 [53.5%] females), 70 300 non−sex-specific cancer cases were documented. In the Chinese population (134 742 participants [54.4% female]), the joint ARD across the 11 risk factors was 182.52 (95% CI, 162.46-202.26) cases per 100 000 person-years, accounting for 84.44% of the male-female incidence rate difference. Smoking (ARD, 111.87 [95% CI, 95.38-128.00]; AP, 51.75%) and heavy alcohol drinking (ARD, 32.64 [95% CI, 24.61-40.43]; AP, 15.10%) were leading contributors. In the UK population (455 024 participants [53.3% female]), the joint ARD was 52.31 (95% CI, 35.63-70.22) cases per 100 000 person-years, accounting for 26.21% of the sex disparity. Obesity (ARD, 15.34 [95% CI, 6.06-24.08]; AP, 7.68%) and smoking (ARD, 12.49 [95% CI, 1.82-24.39]; AP, 6.26%) were leading individual contributors.

Conclusions and Relevance

The findings of this cohort study suggest that modifiable lifestyle factors and health conditions accounted for most of the male-female cancer incidence gap in the Chinese population but only a small fraction in the UK population. Reducing cancer sex disparities may require prevention strategies tailored to population contexts.

This cohort study evaluates the contribution of lifestyle factors and health conditions to sex disparities in cancer incidence across populations.

Introduction

Cancer incidence is consistently higher among males than females for most non−sex-specific sites globally. Although such male predominance is well documented, the underlying drivers remain debated. Biological determinants (eg, genetic makeup, sex hormones, immune responses) contribute to differential susceptibility, while modifiable lifestyle behaviors and health conditions, shaped by societal and cultural norms, may also play a role. Distinguishing their relative contributions is crucial for precision public health interventions.

Despite the importance of this issue, quantitative evidence on how specific risk factors contribute to the male-female cancer gap remains limited. Most previous reports relied on aggregate registry data lacking individual-level exposure, and the few prospective cohort studies addressing this question were restricted to a single population or did not explicitly quantify the contribution of specific risk factors. To our knowledge, no study has systematically quantified and directly compared the contributions of multiple lifestyle factors and health conditions to sex disparities in cancer incidence across diverse populations. The Chinese and UK populations provide an informative contrast in both sociocultural context and risk-factor distribution. Smoking and heavy alcohol drinking are strongly concentrated among males in the Chinese population but are more similar between the sexes in the UK population. Obesity and related metabolic conditions also differ between the settings. These contrasting exposure patterns allow us to assess whether the contribution of modifiable risk factors to cancer disparities by sex varies by population context.

To address this gap, we analyzed individual-level data from more than half a million participants in 3 prospective cohorts in China and the UK, mapping exposure and outcomes to common analytic definitions where possible and applying a unified analytic framework across populations. We quantified sex disparities across 34 non−sex-specific cancer end points and the extent to which 11 lifestyle factors and health conditions contributed to these disparities in diverse population contexts. We hypothesized that modifiable risk factors would account for a larger proportion of the disparity in populations where socially and culturally patterned behaviors, such as smoking and alcohol consumption, differ markedly between males and females. These findings may inform population- and sex-specific cancer prevention.

Methods

The institutional review board of the Shanghai Cancer Institute approved both Chinese studies and the North West Multi-center Research Ethics Committee approved the UK Biobank study. All participants provided written informed consent. Additionally, this work was conducted under the UK Biobank Application No. 833162 and approved by the Renji Hospital Ethics Committee of Shanghai Jiao Tong University School of Medicine (KY2026-016-B). We followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

Study Population

The Shanghai Women’s Health Study (SWHS; 74 940 female residents aged 40-70 years [baseline 1997-2000]) and the Shanghai Men’s Health Study (SMHS; 61 469 male residents aged 40-74 years [baseline 2002-2006]) are population-based prospective cohorts in Shanghai, China. These age ranges reflect the original cohort designs and the concentration of cancer incidence at age 40 years and older. Trained staff conducted in-person baseline interviews to collect information on demographic characteristics, medical and family history, diet, physical activity, and other lifestyle factors. Incident cancer cases and deaths were ascertained through annual linkage to the Shanghai Cancer Registry and Vital Statistics Registry, supplemented by active follow-up. Cancer incidence data were censored on December 31, 2020. Given their shared source population and similar designs, SWHS and SMHS were combined to represent the urban population in Shanghai, China.

The UK Biobank recruited 502 000 participants aged 40 to 70 years from 22 assessment centers across the UK during 2006 to 2010. Touchscreen questionnaires captured sociodemographic factors, lifestyle, family history, and medical history. Cancer diagnoses and deaths were ascertained via linkage to national cancer and death registries. Censoring dates for cancer diagnosis were December 31, 2020 (England), November 30, 2021 (Scotland), and December 31, 2016 (Wales).

Outcome Ascertainment

Cancer outcomes were coded using the International Classification of Diseases, Ninth Revision (ICD-9) for the Chinese cohorts and the ICD-10 for the UK cohort. The study outcome included all non−sex-specific primary cancers (ICD-9 codes 140-195 and 200-208; ICD-10 codes C00-C75 and C81-C96), excluding breast and genital-organ cancers (ICD-9 codes, 174-187; ICD-10 codes, C50-C63) because they affect a single sex either exclusively (genital organs) or almost exclusively (breast), precluding a male-female comparison. End point definitions are detailed in eTable 1 in Supplement 1. We focused on non−sex-specific cancers combined, 4 systems (respiratory, digestive, urinary, hematological), and individual sites with ≥5 incident cases in either sex. Follow-up time was calculated from study enrollment to cancer diagnosis, death, loss to follow-up, or administrative censoring, whichever came first.

Participants with prior cancer at baseline were excluded. Sex-specific cancers during follow-up were not study end points, but affected participants continued contributing person-time. For the combined end point, follow-up was censored at the first non−sex-specific cancer.

Selection of Risk Factors

We selected 11 lifestyle factors and health conditions based on the International Agency for Research on Cancer monographs, the World Cancer Research Fund continuous updating reports, and recent meta-analyses (detailed in eTable 2 in Supplement 1): 3 behavioral factors (cigarette smoking, heavy alcohol drinking, physical inactivity), 4 dietary factors (low intake of vegetable and fruit, high intake of red meat and processed meat), and 4 health conditions (cholelithiasis, chronic liver diseases, diabetes, obesity). Owing to questionnaire heterogeneity across cohorts, exposures with established clinical thresholds were defined accordingly, whereas dietary and physical activity factors without universal cut-points were categorized using population- and sex-specific quartiles (eTable 3 in Supplement 1). Only baseline exposures were used because repeated assessments were unavailable for the entire study population.

Statistical Analysis

Baseline characteristics were summarized by sex within each population and between the 2 populations. Group differences were quantified using standardized differences.

Age-standardized incidence rate (ASIR) per 100 000 person-years (PYs) was calculated for ages 40 to 84 years using the Segi-Doll world standard population. The 95% CIs were calculated based on a gamma distribution to accommodate the small-event-count setting.

Sex disparities were quantified on relative and absolute scales. The male to female incidence rate ratio (IRR) was estimated from a Poisson regression model with log(PYs) as offset, adjusted for age (natural spline, 3 degrees of freedom) and birth cohort (10-year groups). The absolute sex gap was quantified as the difference between the male and female ASIRs, and the 95% CIs were derived from a normal approximation. Age-stratified analyses (5-year strata) were conducted for non−sex-specific cancers combined.

Five-year population attributable fractions (PAFs) for each end point were estimated from sex-specific Cox proportional hazards models including all site-specific risk factors and adjusted for age, birth cohort, cohort-specific categories of education and income, and country within the UK Biobank. Joint PAFs were computed by setting all risk factors to the low-risk reference. Participants with missing covariates were excluded from each model (<1% for all variables except physical activity in UK Biobank [21.3%]; eTable 5 in Supplement 1). “Prefer not to answer” responses for education and income were retained as separate categories.

To integrate rate-based magnitude with risk-factor attribution, we defined the attributable rate difference (ARD) for each factor as ASIR_males × PAF_males − ASIR_females × PAF_females, representing the male-female incidence rate difference (per 100 000 PYs) attributable to that factor (eFigure 1 in Supplement 1). This approach is analogous to established decomposition methods used in prior research. Joint ARDs were computed analogously, since single-factor ARDs under a counterfactual framework are not additive. The 95% CIs were obtained from 1000 nonparametric bootstrap replicates (percentile method). The proportion of the male-female rate difference attributable to the exposures (attributable proportion, AP; %) was calculated as ARD / (ASIR_males − ASIR_females) × 100% (eFigure 1 in Supplement 1).

Three sensitivity analyses were conducted: (1) 10-year PAFs; (2) multiple imputation for physical activity data in the UK Biobank; and (3) Fine-Gray subdistribution hazards models with death as competing event. Mortality from non−sex-specific cancers was analyzed as a secondary outcome using the same framework applied to incidence. Full methodologic details are provided in the eMethods in Supplement 1.

Analyses were performed in R, version 4.5.1 (R Project for Statistical Computing). Two-sided P < .05 was considered statistically significant. Data were analyzed from September 2025 to June 2026.

Results

After excluding participants with prior cancer and mismatched sex, 589 766 participants were included (eFigure 2 in Supplement 1): 134 742 from the 2 Shanghai cohorts (45.6% males) and 455 024 from the UK Biobank (46.7% males). Demographic and risk factor distributions are shown by sex in Table 1, and by study population in eTable 4 in Supplement 1. Because the primary comparisons were within each population, between-population differences were treated as contextual characteristics.

Table 1. Demographic Characteristics and Cancer Risk Factor Distributions by Sex Within Each Population^a.

Characteristic	Sample population, No. (%)
	China			UK
	Male (n = 61 433)	Female (n = 73 309)	SMD	Male (n = 212 520)	Female (n = 242 504)	SMD
Age at study enrollment, median (IQR), y	53.1 (47.4-63.5)	50.3 (44.4-60.9)	0.303	58.1 (50.2-63.7)	57.6 (50.1-63.1)	0.040
Cancer diagnosis	7173 (11.7)	6954 (9.5)	0.071	31 354 (14.8)	24 819 (10.2)	0.025
Incidence rate (95% CI)^b	783.93 (765.78-802.07)	465.09 (454.16-476.02)	NA	1157.62 (1144.81-1170.44)	774.22 (764.59-783.86)	NA
Age at cancer diagnosis, median (IQR), y	58.9 (50.4-68.5)	56.0 (47.2-64.2)	0.371	62.5 (57.5-66.2)	61.7 (56.0-65.6)	0.134
Birth cohort
≤1940	14 570 (23.7)	21 646 (29.5)	0.535	7688 (3.6)	7612 (3.1)	0.054
1941-1950	13 377 (21.8)	19 304 (26.3)		91 435 (43.0)	99 767 (41.1)
1951-1960	26 166 (42.6)	32 346 (44.1)		67 160 (31.6)	81 859 (33.8)
≥1961	7320 (11.9)	13 (0.0)		46 237 (21.8)	53 266 (22.0)
Education level^c
Middle school or above (China); any qualification (UK)	56 489 (92.0)	57 627 (78.6)	0.462	174 435 (82.1)	200 097 (82.5)	0.013
Elementary school or less (China); no formal qualification (UK)	4080 (6.6)	15 669 (21.4)		35 665 (16.8)	39 880 (16.4)
Prefer not to answer	864 (1.4)	13 (0.0)		2420 (1.1)	2527 (1.0)
Income level^d
At or above cohort- and sex-specific threshold	53 584 (87.2)	61 491 (83.9)	0.114	149 659 (70.4)	150 472 (62.0)	0.202
Below cohort- and sex-specific threshold	7721 (12.6)	11 802 (16.1)		37 985 (17.9)	48 014 (19.8)
Prefer not to answer	128 (0.2)	16 (0.0)		24 876 (11.7)	44 018 (18.2)
Ever cigarette smoker^e	42 765 (69.6)	2042 (2.8)	1.935	107 036 (50.7)	95 724 (39.7)	0.222
Heavy alcohol drinker^e^,^f	16 336 (26.6)	663 (0.9)	0.804	53 503 (25.2)	38 628 (15.9)	0.231
Obesity^g	23 232 (37.8)	21 470 (29.3)	0.181	73 320 (34.5)	92 205 (38.0)	0.073
Physical inactivity^h	15 304 (24.9)	18 325 (25.0)	0.002	43 692 (25.0)	45 692 (24.9)	0.001
Not reported	0	0		37 592 (17.7)	59 182 (24.4)
High intake of processed meat^e^,ⁱ	1349 (2.2)	12 484 (17.0)	0.520	91 892 (43.2)	49 990 (20.6)	0.500
High intake of red meat^j	15 358 (25.0)	18 328 (25.0)	<0.001	62 400 (29.4)	60 809 (25.1)	0.096
Low intake of vegetables^k	15 358 (25.0)	18 327 (25.0)	<0.001	44 219 (20.8)	32 323 (13.3)	0.200
Low intake of fruits^k	15 358 (25.0)	18 327 (25.0)	<0.001	19 637 (9.2)	43 542 (18.0)	0.256
Diabetes^e	3861 (6.3)	3167 (4.3)	0.088	15 472 (7.3)	9547 (3.9)	0.146
Chronic liver diseases	4714 (7.7)	1890 (2.6)	0.233	2042 (1.0)	1498 (0.6)	0.039
Cholelithiasis^e	4621 (7.5)	8180 (11.2)	0.125	3548 (1.7)	10 931 (4.5)	0.165

Open in a new tab

Abbreviations: BMI, body mass index (calculated as weight in kilograms divided by height in meters squared); MET, metabolic equivalent of task; NA, not applicable; SMD, standardized mean difference; WC, waist circumference.

^{^a}

The total percentage may not equal 100% due to rounding.

^{^b}

No. of events per 100 000 person-years.

^{^c}

Education definitions were population specific: middle school or above vs elementary school or less in the Chinese population, and any formal qualification vs no formal qualification in the UK population. These categories reflect differences in the original questionnaires and are not directly comparable between populations.

^{^d}

Income definitions were cohort- and sex-specific: annual family income of <¥10 000 (US$1200) vs ≥¥10 000 for Chinese females; personal income of <¥500 ($60)/mo vs ≥¥500/mo for Chinese males; and annual average household income before tax of <£18 000 ($32 000) vs ≥£18 000 for the UK population. US dollar amounts are approximations and were converted using mean annual exchange rates during each cohort’s baseline period and are not adjusted for purchasing power parity. These conversions are provided for reference only and do not imply that these thresholds represent a common poverty standard or an equivalent position in the income distribution. The different income definitions and thresholds were defined in each cohort’s own baseline questionnaire and are not directly comparable across populations or between males and females in the Chinese cohorts.

^{^e}

Variables with missing values (proportion <1%). Details are presented in eTable 5 in Supplement 1. Proportions were calculated based on participants with nonmissing values.

^{^f}

Chinese population: alcohol intake frequency ≥7 times per week; UK population: alcohol intake frequency: daily or almost daily.

^{^g}

Chinese population: BMI ≥28; WC ≥83 cm (females) or 88 cm (males); UK population: BMI≥30 kg/m²; WC≥88 cm (females) or 102 cm (males).

^{^h}

Level of total physical activity (MET-h/week for Chinese population and MET-min/week for UK population)<the 25th percentile. Proportions were calculated based on participants with nonmissing values.

^ⁱ

Frequency of processed meat intake >2 times (Chinese population) or ≥2 times (UK population) per week.

^{^j}

Intake ≥75th percentile.

^{^k}

Intake <25th percentile.

Sex Disparities in Cancer Incidence

During follow-up, 70 300 incident non−sex-specific cancer cases were documented. ASIRs are shown in eFigure 3 in Supplement 1. The overall male to female IRR was higher in the Chinese population (1.66; 95% CI, 1.60-1.72) than in the UK population (1.49; 95% CI, 1.46-1.51) (Table 2). On the absolute scale, the differences in ASIRs were 216.15 (95% CI, 197.37-234.93) and 199.62 (95% CI, 184.92-214.32) cases per 100 000 PYs (Table 2). The male-female gap emerged at an earlier age in the Chinese population than in the UK population (eResults and eFigures 4 and 5 in Supplement 1). Mortality was examined as a secondary and descriptive outcome and the results are presented in the eResults, eFigures 5 to 8, and eTables 7 and 9 in Supplement 1.

Table 2. Male to Female Incidence Rate Ratios (IRRs) and Differences in Age-Standardized Incidence Rates (ASIRs) by Cancer Site.

Cancer site	Chinese population		UK population
Cancer site	IRR (95% CI)	Difference in ASIRs (95% CI)	IRR (95% CI)	Difference in ASIRs (95% CI)
Non−sex-specific cancers combined	1.66 (1.60 to 1.72)	216.15 (197.37 to 234.93)	1.49 (1.46 to 1.51)	199.62 (184.92 to 214.32)
Respiratory system	1.80 (1.68 to 1.92)	71.99 (62.39 to 81.60)	1.35 (1.28 to 1.43)	13.13 (9.75 to 16.51)
Nasal cavities, middle ear and accessory sinuses	2.82 (1.49 to 5.35)	2.44 (0.18 to 4.70)	1.34 (0.82 to 2.18)	1.05 (−0.04 to 2.13)
Larynx	13.98 (7.45 to 26.22)	9.12 (5.90 to 12.33)	11.82 (7.47 to 18.71)	3.83 (3.15 to 4.51)
Lung	1.71 (1.60 to 1.84)	59.91 (51.23 to 68.58)	1.26 (1.19 to 1.34)	8.07 (4.96 to 11.19)
Thymus, heart, and mediastinum	1.83 (1.03 to 3.25)	0.65 (−0.49 to 1.80)	1.10 (0.61 to 2.00)	0.05 (−0.19 to 0.29)
Digestive system	1.75 (1.66 to 1.83)	111.71 (99.09 to 124.33)	1.63 (1.57 to 1.69)	50.00 (44.02 to 55.99)
Esophagus	3.93 (3.02 to 5.12)	12.42 (8.93 to 15.90)	3.13 (2.75 to 3.55)	11.67 (10.10 to 13.24)
Stomach	2.29 (2.07 to 2.53)	35.60 (28.58 to 42.62)	2.35 (2.04 to 2.72)	6.82 (5.19 to 8.46)
Small intestine	1.13 (0.74 to 1.73)	0.34 (−0.78 to 1.46)	1.45 (1.16 to 1.82)	1.69 (0.46 to 2.92)
Colon	1.32 (1.21 to 1.44)	11.36 (6.08 to 16.64)	1.28 (1.21 to 1.36)	7.66 (3.79 to 11.54)
Rectum	1.66 (1.48 to 1.87)	14.49 (9.50 to 19.48)	1.91 (1.76 to 2.07)	13.57 (10.65 to 16.49)
Liver	2.83 (2.47 to 3.24)	33.23 (27.84 to 38.63)	2.46 (2.11 to 2.87)	5.46 (4.30 to 6.63)
Gallbladder	0.82 (0.66 to 1.02)	−2.70 (−4.86 to −0.54)	1.09 (0.89 to 1.33)	0.54 (−0.48 to 1.55)
Pancreas	1.48 (1.28 to 1.71)	7.85 (4.80 to 10.90)	1.35 (1.22 to 1.49)	4.02 (2.05 to 6.00)
Other and ill-defined digestive organs	1.63 (0.80 to 3.33)	0.21 (−0.32 to 0.73)	0.98 (0.52 to 1.83)	−0.03 (−0.32 to 0.26)
Urinary system	2.59 (2.28 to 2.93)	32.30 (26.64 to 37.97)	2.78 (2.58 to 3.00)	27.91 (25.21 to 30.61)
Kidney	1.91 (1.62 to 2.25)	15.41 (11.30 to 19.52)	2.27 (2.07 to 2.49)	14.05 (11.79 to 16.32)
Bladder	3.89 (3.18 to 4.75)	16.94 (13.03 to 20.84)	4.01 (3.53 to 4.56)	14.41 (12.92 to 15.89)
Hematologic system	1.43 (1.25 to 1.65)	7.21 (3.17 to 11.24)	1.51 (1.43 to 1.60)	22.85 (18.25 to 27.45)
Non-Hodgkin lymphoma	1.26 (1.04 to 1.52)	2.74 (−0.11 to 5.60)	1.47 (1.38 to 1.57)	14.97 (11.38 to 18.56)
Hodgkin disease	1.55 (0.50 to 4.84)	0.11 (−0.28 to 0.50)	1.40 (1.05 to 1.88)	0.53 (−0.68 to 1.75)
Leukemia	1.60 (1.22 to 2.10)	2.54 (0.15 to 4.93)	1.62 (1.38 to 1.90)	2.55 (1.04 to 4.05)
Multiple myeloma	1.82 (1.30 to 2.54)	1.76 (0.26 to 3.26)	1.62 (1.44 to 1.82)	5.22 (3.07 to 7.38)
Other systems
Oral cavity	2.35 (1.89 to 2.94)	9.22 (6.11 to 12.33)	2.29 (2.03 to 2.58)	12.17 (9.92 to 14.41)
Thyroid	0.33 (0.27 to 0.40)	−19.82 (−24.20 to −15.43)	0.45 (0.37 to 0.54)	−5.67 (−7.67 to −3.66)
Other endocrine glands and related structures	1.17 (0.56 to 2.48)	0.10 (−0.57 to 0.77)	1.16 (0.61 to 2.19)	0.41 (−0.38 to 1.19)
Melanoma	0.78 (0.36 to 1.66)	−0.47 (−1.38 to 0.44)	1.25 (1.16 to 1.33)	0.40 (−4.38 to 5.19)
Nonmelanoma skin	1.36 (1.03 to 1.80)	2.07 (0.27 to 3.87)	1.43 (1.40 to 1.46)	80.84 (70.97 to 90.71)
Central nervous system	0.84 (0.67 to 1.07)	−1.04 (−4.75 to 2.67)	1.62 (1.42 to 1.85)	5.77 (3.52 to 8.02)
Heart, mediastinum, and pleura	1.80 (0.51 to 6.37)	0.15 (−0.22 to 0.52)	2.56 (0.89 to 7.36)	0.49 (−0.25 to 1.22)
Retroperitoneum and peritoneum	0.53 (0.25 to 1.14)	−0.55 (−1.39 to 0.30)	0.20 (0.13 to 0.30)	−2.41 (−3.27 to −1.55)
Bone and articular cartilage	1.01 (0.53 to 1.93)	0.19 (−0.63 to 1.01)	2.19 (1.27 to 3.76)	0.95 (0.09 to 1.81)
Connective and other soft tissue	1.39 (0.85 to 2.28)	−0.39 (−1.86 to 1.08)	1.77 (1.39 to 2.25)	1.30 (−0.27 to 2.87)

Open in a new tab

Abbreviations: IRR, incidence rate ratio; ASIR, age-standardized incidence rate.

The IRR was estimated from a Poisson regression model adjusted for age and birth cohort.

Sex-Specific Population Attributable Fractions

The 5-year PAFs of 11 risk factors are summarized in Figure 1. In the Chinese population, smoking was the dominant individual factor in males, with PAFs of 21.67% for non−sex-specific cancers combined and 57.61% for lung cancer, vs 0.79% and 4.04% in females (Figure 1A). Heavy alcohol drinking had particularly high PAFs for cancers of esophagus, oral cavity, and stomach. Conversely, obesity had higher PAFs among Chinese females for some cancers (eg, liver, 19.98% vs 2.04%).

Two-panel heatmap table of cancer P A F by risk factor, sex, and site. Two stacked panels labeled A Chinese population and B U K population. Each panel is a heatmap table with rows of risk factors at left and columns of cancer sites across the top; each site is split into M and F subcolumns. Column headers, left to right: All, Lung, Esophagus, Stomach, Colon, Rectum, Liver, Gallbladder, Pancreas, Kidney, Bladder, Oral cavity, Thyroid. Row labels, top to bottom: Cigarette smoking; Alcohol drinking; Physical inactivity; Decreased vegetable intake; Decreased fruit intake; Increased red meat intake; Increased processed meat intake; Obesity; Type 2 diabetes; Chronic liver diseases; Cholelithiasis; Joint (all factors). Each cell contains a numeric P A F value in percent, with a diverging color fill from purple for negative values through white near zero to orange and red for higher positive values; some cells are blank. In panel A, notable high positive cells include Lung M 57.61, Esophagus M 35.54 and Alcohol drinking Esophagus M 46.09, Liver M 17.00 with Chronic liver diseases Liver M 52.02, Gallbladder M 23.21 and F 27.08 for Cholelithiasis, and Joint row values including Lung M 67.02, Esophagus M 68.80, Liver M 65.61, Gallbladder M 42.42. Negative examples include Thyroid M minus 10.76 for Cigarette smoking and Esophagus F minus 9.87 for Physical inactivity. In panel B, high positive cells include Cigarette smoking Lung M 75.51 and F 66.63, Bladder M 38.62, Oral cavity M 34.11 and F 27.90; Obesity Liver M 28.01; Type 2 diabetes Liver M 27.48; Chronic liver diseases Liver M 24.94; Joint row values including Lung M 82.53 and F 72.72, Liver M 70.70 and F 44.78, Oral cavity M 45.71 and F 26.75. A legend at the bottom indicates Sex colors for Female and Male, category of risk factors as Behavioral, Dietary, and Health condition, and system of cancer site as All, Respiratory, Digestive, Urinary, and Other. A horizontal color bar labeled P A F, percent ranges from minus 10 to 80. — Bolded values indicate statistical significance (P < .05); blank cells are those in which the risk factor was not modeled for the cancer site; and “All” is non−sex-specific cancers combined.

In the UK population, the contribution of risk factors was more similar across sexes (Figure 1B). Although smoking prevalence was higher in males, the lung-cancer PAF for smoking was substantial in both sexes (75.51% in males, 66.63% in females). Obesity was prominent in UK females for gallbladder cancer (17.08% vs 6.11% in males) and thyroid cancer (12.56% vs 3.81%), contrasting with the Chinese pattern.

Risk Factor Contributions to the Male-Female Cancer Rate Gap

We translated PAFs into ARDs and APs, quantifying the number and proportion of the male-female incidence gap attributable to the modeled risk factors (Figures 2 and 3; eTable 8 in Supplement 1; conceptual schematic and worked example in eFigure 1 in Supplement 1). For non−sex-specific cancers combined, the joint ARD across the 11 risk factors was 182.52 (95% CI, 162.46-202.26) per 100 000 PYs in the Chinese population, accounting for 84.44% of the sex gap. In the UK population, the corresponding joint ARD and AP were 52.31 (95% CI, 35.63-70.22) and 26.21%. Because single-factor ARDs and APs are not additive, joint and single-factor estimates are presented separately (Figures 2 and 3). The ARD estimates are quantitative attributions under the assumptions of the underlying Cox models and do not represent definitive causal partitions.

Two-panel bar charts of joint A R D by cancer site for China and U K. Two stacked panels labeled A and B, each combining a left table and a right horizontal bar plot. Panel A title at upper left: Chinese population. Panel B title at upper left: U K population. In both panels, the left table has column header Cancer site and column header Joint A R D (95% C I) per 100000 P Y s. Rows list site groups and sites. Panel A table values: Non-sex-specific cancers combined 182.52 (162.46 to 202.26); Lung 82.56 (73.92 to 90.44); Esophagus 10.35 (7.42 to 13.72); Stomach 17.30 (10.29 to 23.82); Colon 11.56 (4.33 to 18.67); Rectum 14.60 (8.79 to 20.29); Liver 26.10 (21.82 to 30.82); Gallbladder minus 0.27 (minus 2.01 to 1.36); Pancreas 3.75 (minus 0.20 to 7.17); Kidney 1.23 (minus 3.71 to 5.96); Bladder 6.55 (2.96 to 10.33); Oral cavity 2.13 (minus 2.01 to 6.13); Thyroid minus 1.48 (minus 6.66 to 3.05). Panel B table values: Non-sex-specific cancers combined 52.31 (35.63 to 70.22); Lung 10.37 (7.36 to 13.33); Esophagus 8.02 (6.39 to 9.73); Stomach 3.38 (1.88 to 4.80); Colon 6.62 (2.55 to 10.77); Rectum 6.11 (2.93 to 9.08); Liver 4.97 (3.90 to 6.06); Gallbladder minus 0.06 (minus 0.67 to 0.62); Pancreas 1.94 (0.24 to 3.67); Kidney 4.25 (2.25 to 6.14); Bladder 6.99 (5.50 to 8.49); Oral cavity 7.12 (5.18 to 9.07); Thyroid minus 0.80 (minus 2.75 to 1.11). To the right of each table, light blue horizontal bars align with the same row order; thin black whiskers mark 95 percent confidence intervals. A vertical dotted line marks zero. Text above the bars reads Contribute to higher rates in female individuals on the left of zero and Contribute to higher rates in male individuals on the right. The horizontal axis label reads Joint A R D (95% C I) per 100000 P Y s with tick labels at approximately minus 5, 0, 5, 10, 25, 50, 100, and 200. A rightmost column labeled A P, percent lists, for Panel A: 84.44, 137.82, 83.36, 48.61, 101.70, 100.76, 78.55, 9.96, 47.77, 7.97, 38.66, 23.05, 7.45; for Panel B: 26.21, 128.43, 68.72, 49.50, 86.35, 45.01, 90.99, minus 10.68, 48.14, 30.24, 48.53, 58.56, 14.09. — ARD is defined as ASIR_male × PAF_male − ASIR_female × PAF_female and is interpretable as the number of cases per 100 000 PYs of the male-female incidence rate gap attributable to the modeled risk factors. AP is defined as ARD / (ASIR_male − ASIR_female) and is interpretable as the proportion of the male-female incidence rate gap attributable to the modeled risk factors. A conceptual schematic and a worked example are provided in eFigure 1 in Supplement 1. The joint ARD and AP reflect the combined, jointly estimated contribution of all risk factors modeled for a given site. Bars extending to the right (positive ARD) indicate a contribution to higher rates in males, and bars extending to the left (negative ARD) indicate a contribution to higher rates in females. Error bars denote 95% CIs. An interval crossing zero indicates the estimate was not statistically significant at the α = .05 level. Because single-factor ARDs are not additive, joint ARDs are presented separately, whereas single-factor ARDs are shown in Figure 3. The specific set of risk factors combined to compute the joint ARD and AP for each cancer site is provided in eTable 2 in Supplement 1. ASIR indicates age-standardized incidence rate; PAF, population attributable fraction; PYs, person-years.

Two-panel circular bar charts comparing A R D by risk factor for China and U K. Two circular bar plots stacked vertically, labeled panel A Chinese population at the upper left of the top plot and panel B U K population at the upper left of the bottom plot. Each plot is a ring of colored sector headers around the outside with inward and outward radial bars inside the ring. The outer sector labels, printed along the arc, include cigarette smoking, alcohol drinking, physical inactivity, decreased vegetable, decreased fruit, increased red meat, increased processed meat, obesity, type 2 diabetes, chronic liver disease, and cholelithiasis. Within each sector, cancer site labels repeat around the circle in black text, including All, lung, esophagus, stomach, colon, rectum, liver, gallbladder, pancreas, kidney, bladder, thyroid, and oral cavity; the exact set varies by sector but these labels are visible multiple times around the ring. A radial scale at the upper left of each plot is labeled A R D with dotted circular gridlines and tick labels at approximately 0, 2, 5, 10, 20, 50, and 100. Bars extend radially from near the inner white center toward the dotted gridlines; some bars extend farther outward than others. Bars are colored to match their sector theme and many additional bars are light gray. In both panels, the cigarette smoking sector contains several of the tallest outward bars, reaching near the 50 to 100 rings for some cancer labels. Alcohol drinking and physical inactivity sectors contain shorter bars clustered mostly below about 20. Diet-related sectors for decreased vegetable, decreased fruit, increased red meat, and increased processed meat contain mixed short to moderate bars, with several gray bars interspersed. On the right side of each plot, the obesity sector contains multiple medium to tall brown bars, with one prominent bar extending well beyond neighboring bars. The type 2 diabetes sector contains several orange bars of small to moderate height. The chronic liver disease sector contains green bars including one noticeably taller bar compared with adjacent green bars. The cholelithiasis sector at the top right contains a few short yellow-green bars near the inner rings. No legend box is present; color is indicated by the outer arc labels. — ARD is defined as ASIR_male × PAF_male − ASIR_female × PAF_female, and is interpretable as the absolute number of cases per 100 000 person-years of the male-female incidence rate gap attributable to a given risk factor. Bars extending outwards (positive ARD) indicate that the factor contributes to higher rates in males. Bar height corresponds to the absolute magnitude of the ARD. Bars in gray denote ARDs whose 95% CIs crossed zero (ie, not statistically significant at the α = .05 level). Because single-factor ARDs are not additive, their combined (joint) contribution across all modeled factors is shown separately in Figure 2. A conceptual schematic and a worked example are provided in eFigure 1 in Supplement 1. ARDs are model-based attributions and do not represent definitive causal partitions. All indicates all the non−sex-specific cancers combined; ASIR, age-standardized incidence rate; PAF, population attributable fraction.

In the Chinese population, the male-female incidence gap was dominated by behavioral factors of large magnitude (Figure 3A; eTable 8 in Supplement 1). Cigarette smoking alone was associated with 51.75% of the sex gap (ARD, 111.87 [95% CI, 95.38-128.00]) and was the leading contributor to the male-female gap in lung, stomach, rectum, and liver cancers. Heavy alcohol drinking exhibited an ARD of 32.64 (95% CI, 24.61-40.43) cases per 100 000 PYs and AP of 15.10%, with contributions concentrated in lung, esophagus, and stomach cancers. Chronic liver diseases (overall ARD, 21.72 [95% CI, 17.68-26.24]; AP, 10.05%) and low fruit intake (overall ARD, 21.42 [95% CI, 11.53-31.42]; AP, 9.91%) were the next leading contributors. Metabolic factors, including obesity and diabetes, contributed minimally to the gap for non−sex-specific cancers combined.

In the UK population, the contributing factors were more heterogeneous and were dominated by metabolic rather than behavioral exposures (Figure 3B; eTable 8 in Supplement 1). For non−sex-specific cancers combined, obesity was the leading contributor (ARD, 15.34 [95% CI, 6.06-24.08]; AP, 7.68%), followed by smoking (ARD, 12.49 [95% CI, 1.82-24.39]; AP, 6.26%) and diabetes (ARD, 7.00 [95% CI, 3.39-10.56]; AP, 3.51%).

Sensitivity Analyses

Extending the PAF horizon from 5 to 10 years yielded estimates concordant with the primary analysis, in both direction and magnitude (eTable 8 in Supplement 1). Subdistribution-hazard estimates (Fine-Gray model, with death as competing event) were materially similar to cause-specific Cox estimates (eTable 10 in Supplement 1). Multiple imputation for total physical activity in the UK Biobank (21.3% missing; eTables 5 and 6 in Supplement 1) yielded PAFs nearly identical to the complete-case analysis (eTable 11 in Supplement 1). Details of the sensitivity analyses are provided in eResults in Supplement 1.

Discussion

In 2 populations with contrasting risk factor profiles, we found that both the magnitude and composition of modifiable contributions to cancer sex disparities were highly setting-dependent. In the Chinese population, the male-female gap was predominantly attributable to marked sex differences in lifestyle behaviors, particularly cigarette smoking and heavy alcohol drinking. In contrast, in the UK population, these factors accounted for a substantially smaller proportion of the sex gap, leaving much of the disparity unexplained by the measured factors.

Although male predominance in cancer is consistent globally, the magnitude varies across sites, regions, and time periods. Registry-based studies have inferred potential drivers for this variation. For example, Carioli et al attributed epithelial cancer mortality patterns to historical smoking and alcohol epidemics, while Edgren et al labeled 13 cancer sites as “enigmatic” for lacking obvious environmental explanations. Our study extends these observations by shifting from ecological inference to individual-level quantification, expressing each factor’s contribution as absolute rate difference and a proportion of the male-female rate gap. Our findings in the UK population align with other Western cohorts. The US National Institutes of Health−American Association of Retired Persons study, using Peters-Belson decomposition (analogous to our ARD framework), found that smoking, alcohol, anthropometrics, diet, medication use, and medical history accounted for only a modest proportion (11.2%-49.4%) of the male-female incidence gap at 7 cancer sites and none at 12 sites. Surveillance, Epidemiology, and End Results program (US National Cancer Institute) analyses across 5 US racial and ethnic groups likewise showed largely stable male-to-female incidence rate ratios despite substantial differences in smoking and obesity prevalence, suggesting that lifestyle variation alone is insufficient to close the sex gap in Western populations. Our Chinese data indicate that this modest contribution is not universal. Where lifestyle patterns differ markedly between males and females, a substantial portion of the sex disparity in cancer is attributable to modifiable behaviors and therefore potentially preventable.

This context-dependent dichotomy is particularly evident in alcohol-related carcinogenesis. In East Asian populations, aldehyde dehydrogenase 2 (ALDH2) deficiency is common and limits acetaldehyde detoxification, amplifying cancer risk, while sociocultural norms are associated with much lower alcohol consumption among females. Therefore, alcohol exposure and its interaction with ALDH2 deficiency is concentrated predominantly in males. Targeted alcohol control among males may be a high-yield strategy for reducing sex disparities in alcohol-related cancers.

Metabolic dysregulation also contributed heterogeneously. In the UK population, diabetes contributed to higher incidence of liver and pancreatic cancers in males, highlighting the potential role of glycemic control. Conversely, obesity PAFs for gallbladder and thyroid cancers was concentrated in Chinese males but higher among UK females. This divergence indicates that obesity-attributable patterns are not uniform across populations. Adiposity promotes tumorigenesis through competing mechanisms, including insulin resistance, chronic inflammation, and sex hormone metabolism. Hormone-mediated pathways may predominate in the UK population, whereas inflammatory pathways linked to visceral adiposity, to which East Asian males are more predisposed, may dominate in the Chinese population. These hypotheses warrant investigation using body-composition and biomarker data.

Chronic liver disease contributed substantially to liver-cancer sex disparity in both populations, likely through distinct causes. Hepatitis B virus−related hepatitis and cirrhosis likely predominate in the Chinese population, whereas metabolic dysfunction-associated steatotic liver disease may be more prominent in the UK population, consistent with the greater contributions of obesity and diabetes to liver cancer that we observed.

Limitations

First, as an observational study, residual confounding and reverse causation cannot be ruled out. Because the SWHS and SMHS enrolled participants during slightly different periods (1997-2000 vs 2002-2006), sex is structurally confounded with cohort and calendar time in the Chinese analysis. Given rapid lifestyle transitions in China during this time, residual confounding by calendar time may persist despite adjustment for age and birth cohort. Income adjustment was crude and based on measures that were noncomparable across populations and internally inconsistent between the 2 Chinese cohorts. Education definitions also differed across populations. Although these variables were used only as covariates in sex- and population-specific models, they may have resulted in residual or differential socioeconomic confounding. Second, risk factors were assessed only at baseline to maintain data harmonization, and subsequent changes were not captured. Together with self-reporting, this may have introduced exposure misclassification. Third, we lacked data on occupational exposures, ambient air pollution, and specific infectious agents (eg, Helicobacter pylori, Epstein-Barr virus), precluding direct assessment of their contributions. These factors may be correlated with measured behaviors, and the estimated ARDs for behavioral factors may partially capture their effects. Fourth, dietary and physical activity factors were categorized using population- and sex-specific quartiles and therefore reflect relative rather than absolute exposure. Fifth, screening-coverage differences between the UK and China may have influenced site-specific incidence and ARDs. Sixth, all cohorts enrolled participants aged 40 to 74 years at baseline. Although follow-up extended into older ages, the oldest groups were underrepresented and early-onset cancers (age <40 years) were not captured. Because both the magnitude of the male-female gap and the contribution of specific factors may differ at the extremes of age, our estimates may not generalize to these groups and warrant investigation in cohorts spanning a wider age range. Finally, generalizability is limited because the Shanghai cohorts represent an urban Chinese population, whereas the UK Biobank is subject to a well-documented healthy-volunteer selection bias and is predominantly of White European ancestry. Race and ethnicity were unavailable in the Chinese cohorts, so within-population ethnic heterogeneity could not be examined. Therefore, comparisons across racial and ethnic populations should be interpreted with caution.

Conclusions

In this cohort study, contributions of lifestyle behaviors and health conditions to cancer sex disparities were not uniform across populations. In the Chinese population, a potential priority for helping address the cancer disparity by sex could be in targeting behavioral interventions, specifically tobacco and alcohol control among males. In the UK population, the measured risk factors accounted for only a modest proportion of the male-female rate gap, underscoring the need to investigate unmeasured occupational and environmental exposures, as well as intrinsic biological determinants. Recognizing that biological sex and sociocultural gender contribute differently to cancer across populations is essential for global cancer equity and sex-specific precision prevention strategies.

Supplement 1.

eMethods. Detailed Statistical Methodology

eResults. Supplementary Results

eFigure 1. Conceptual Framework of the Attributable Rate Difference (ARD) with A Worked Example

eFigure 2. Study Population and Flow Diagram

eFigure 3. Sex-Specific Age-Standardized Incidence Rates of Cancer

eFigure 4. Sex Disparities for the Incidence of Non−sex-specific Cancers Combined (by 5-Year Age Group)

eFigure 5. Sex-Specific Cumulative Incidence and Mortality of Non−sex-specific Cancers Combined (40-84 Years)

eFigure 6. Sex-Specific Age-Standardized Mortality Rates of Cancer

eFigure 7. Sex Disparities for the Mortality of Non−sex-specific Cancers Combined (by 5-Year Age Group)

eFigure 8. Male-to-Female Incidence Rate Ratios and Mortality Rate Ratios by Cancer Site

eTable 1. ICD-9 and ICD-10 Codes for Non−sex-specific Cancer Types and Number of New Cases and Deaths for Each Outcome

eTable 2. Risk Factors Used to Calculate Population Attributable Fractions for Each Cancer Site eTable 3. Summary of Risk Factors Used to Calculate Population Attributable Fractions

eTable 4. Demographics and Distributions of Risk Factors, by Study Population

eTable 5. Missing Data and Follow-Up Summary by Population and Sex

eTable 6. Comparison of UKB Participants With and Without Complete Data on Total Physical Activity

eTable 7. Male-to-Female Mortality Rate Ratios and Differences in Age-standardized Mortality Rate by Cancer Site

eTable 8. 5-year and 10-year Population Attributable Fractions (PAFs), Attributable Rate Differences (ARDs) and Attributable Proportion (AP) for Cancer Incidence

eTable 9. 5-year Population Attributable Fractions (PAFs) and Attributable Rate Differences (ARDs) for Cancer Mortality

eTable 10. Cause-Specific Hazard Ratios (HRs) and Fine-Gray Subdistribution Hazard Ratios (sdHRs) for Associations between Risk Factors and Cancer Outcomes

eTable 11. 5-year Population Attributable Fractions under Complete-Case Analysis vs Multiple Imputation for Total Physical Activity (UK Biobank)

jamaoncol-e263747-s001.pdf^{(12.1MB, pdf)}

Supplement 2.

Data Sharing Statement

jamaoncol-e263747-s002.pdf^{(16.3KB, pdf)}

References

1.Bray F, Laversanne M, Sung H, et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2024;74(3):229-263. doi: 10.3322/caac.21834 [DOI] [PubMed] [Google Scholar]
2.Selvaraj RC, Cioffi G, Waite KA, Jackson SS, Barnholtz-Sloan JS. A pan-cancer analysis of age and sex differences in cancer incidence and survival in the United States, 2001-2020. Cancers (Basel). 2025;17(3):378. doi: 10.3390/cancers17030378 [DOI] [PMC free article] [PubMed] [Google Scholar]
3.Hoveling LA, Eijkelboom A, Schuurman M, et al. Disparities in cancer incidence and stage at diagnosis between male and female patients in the Netherlands. Eur J Cancer. 2025;230:116063. doi: 10.1016/j.ejca.2025.116063 [DOI] [PubMed] [Google Scholar]
4.Tosakoon S, Lawrence WR, Shiels MS, Jackson SS. Sex differences in cancer incidence rates by race and ethnicity: Results from the surveillance, epidemiology, and end results (SEER) registry (2000-2019). Cancers (Basel). 2024;16(5):989. doi: 10.3390/cancers16050989 [DOI] [PMC free article] [PubMed] [Google Scholar]
5.Radkiewicz C, Bruchfeld JB, Weibull CE, et al. Sex differences in lymphoma incidence and mortality by subtype: a population-based study. Am J Hematol. 2023;98(1):23-30. doi: 10.1002/ajh.26744 [DOI] [PMC free article] [PubMed] [Google Scholar]
6.Al-Ani A, Amarin JZ, Al-Huneidy L, Al-Hussaini M. Sex differences in cancer incidence and mortality among patients managed at King Hussein Cancer Center. Int J Cancer. 2022;151(11):1960-1968. doi: 10.1002/ijc.34213 [DOI] [PubMed] [Google Scholar]
7.Carioli G, Bertuccio P, Levi F, et al. Cohort analysis of epithelial cancer mortality male-to-female sex ratios in the European Union, USA, and Japan. Int J Environ Res Public Health. 2020;17(15):5311. doi: 10.3390/ijerph17155311 [DOI] [PMC free article] [PubMed] [Google Scholar]
8.Dong M, Cioffi G, Wang J, et al. Sex differences in cancer incidence and survival: a pan-cancer analysis. Cancer Epidemiol Biomarkers Prev. 2020;29(7):1389-1397. doi: 10.1158/1055-9965.EPI-20-0036 [DOI] [PubMed] [Google Scholar]
9.Radkiewicz C, Johansson ALV, Dickman PW, Lambe M, Edgren G. Sex differences in cancer risk and survival: a Swedish cohort study. Eur J Cancer. 2017;84:130-140. doi: 10.1016/j.ejca.2017.07.013 [DOI] [PubMed] [Google Scholar]
10.Edgren G, Liang L, Adami HO, Chang ET. Enigmatic sex disparities in cancer incidence. Eur J Epidemiol. 2012;27(3):187-196. doi: 10.1007/s10654-011-9647-5 [DOI] [PubMed] [Google Scholar]
11.Cook MB, Dawsey SM, Freedman ND, et al. Sex disparities in cancer incidence by period and age. Cancer Epidemiol Biomarkers Prev. 2009;18(4):1174-1182. doi: 10.1158/1055-9965.EPI-08-1118 [DOI] [PMC free article] [PubMed] [Google Scholar]
12.Zheng D, Trynda J, Williams C, et al. Sexual dimorphism in the incidence of human cancers. BMC Cancer. 2019;19(1):684. doi: 10.1186/s12885-019-5902-z [DOI] [PMC free article] [PubMed] [Google Scholar]
13.Khan M, Papier K, Pirie KL, Key TJ, Atkins J, Travis RC. Sex differences in cancer incidence: prospective analyses in the UK Biobank. Br J Cancer. 2025;133(2):216-226. doi: 10.1038/s41416-025-03028-y [DOI] [PMC free article] [PubMed] [Google Scholar]
14.Mauvais-Jarvis F, Bairey Merz N, Barnes PJ, et al. Sex and gender: modifiers of health, disease, and medicine. Lancet. 2020;396(10250):565-582. doi: 10.1016/S0140-6736(20)31561-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
15.Rubin JB. Gender and sex interactions are intrinsic components of cancer phenotypes. Nat Rev Cancer. 2025;25(8):634-648. doi: 10.1038/s41568-025-00829-4 [DOI] [PubMed] [Google Scholar]
16.Fink H, Langselius O, Vignat J, et al. Global and regional cancer burden attributable to modifiable risk factors to inform prevention. Nat Med. Published online February 3, 2026:1-10. doi: 10.1038/s41591-026-04219-7 [DOI] [PubMed] [Google Scholar]
17.Jackson SS, Marks MA, Katki HA, et al. Sex disparities in the incidence of 21 cancer types: quantification of the contribution of risk factors. Cancer. 2022;128(19):3531-3540. doi: 10.1002/cncr.34390 [DOI] [PMC free article] [PubMed] [Google Scholar]
18.GBD 2019 Risk Factors Collaborators . Global burden of 87 risk factors in 204 countries and territories, 1990-2019: a systematic analysis for the Global Burden of Disease Study 2019. Lancet. 2020;396(10258):1223-1249. doi: 10.1016/S0140-6736(20)30752-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
19.Uddin R, Lee EY, Khan SR, Tremblay MS, Khan A. Clustering of lifestyle risk factors for non-communicable diseases in 304,779 adolescents from 89 countries: a global perspective. Prev Med. 2020;131:105955. doi: 10.1016/j.ypmed.2019.105955 [DOI] [PubMed] [Google Scholar]
20.Reitsma MB, Kendrick PJ, Ababneh E, et al. ; GBD 2019 Tobacco Collaborators . Spatial, temporal, and demographic patterns in prevalence of smoking tobacco use and attributable disease burden in 204 countries and territories, 1990-2019: a systematic analysis from the Global Burden of Disease Study 2019. Lancet. 2021;397(10292):2337-2360. doi: 10.1016/S0140-6736(21)01169-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
21.Bryazka D, Reitsma MB, Griswold MG, et al. ; GBD 2020 Alcohol Collaborators . Population-level risks of alcohol consumption by amount, geography, age, sex, and year: a systematic analysis for the Global Burden of Disease Study 2020. Lancet. 2022;400(10347):185-235. doi: 10.1016/S0140-6736(22)00847-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
22.Phelps NH, Singleton RK, Zhou B, et al. ; NCD Risk Factor Collaboration (NCD-RisC) . Worldwide trends in underweight and obesity from 1990 to 2022: a pooled analysis of 3663 population-representative studies with 222 million children, adolescents, and adults. Lancet. 2024;403(10431):1027-1050. doi: 10.1016/S0140-6736(23)02750-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
23.Zhou B, Rayner AW, Gregg EW, et al. ; NCD Risk Factor Collaboration (NCD-RisC) . Worldwide trends in diabetes prevalence and treatment from 1990 to 2022: a pooled analysis of 1108 population-representative studies with 141 million participants. Lancet. 2024;404(10467):2077-2093. doi: 10.1016/S0140-6736(24)02317-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
24.Chen W, Xia C, Zheng R, et al. Disparities by province, age, and sex in site-specific cancer burden attributable to 23 potentially modifiable risk factors in China: a comparative risk assessment. Lancet Glob Health. 2019;7(2):e257-e269. doi: 10.1016/S2214-109X(18)30488-1 [DOI] [PubMed] [Google Scholar]
25.Brown KF, Rumgay H, Dunlop C, et al. The fraction of cancer attributable to modifiable risk factors in England, Wales, Scotland, Northern Ireland, and the United Kingdom in 2015. Br J Cancer. 2018;118(8):1130-1141. doi: 10.1038/s41416-018-0029-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
26.Zheng W, Chow WH, Yang G, et al. The Shanghai Women’s Health Study: rationale, study design, and baseline characteristics. Am J Epidemiol. 2005;162(11):1123-1131. doi: 10.1093/aje/kwi322 [DOI] [PubMed] [Google Scholar]
27.Shu XO, Li H, Yang G, et al. Cohort profile: the Shanghai Men’s Health Study. Int J Epidemiol. 2015;44(3):810-818. doi: 10.1093/ije/dyv013 [DOI] [PMC free article] [PubMed] [Google Scholar]
28.Shu XO, Yang G, Jin F, et al. Validity and reproducibility of the food frequency questionnaire used in the Shanghai Women’s Health Study. Eur J Clin Nutr. 2004;58(1):17-23. doi: 10.1038/sj.ejcn.1601738 [DOI] [PubMed] [Google Scholar]
29.Villegas R, Yang G, Liu D, et al. Validity and reproducibility of the food-frequency questionnaire used in the Shanghai Men’s Health Study. Br J Nutr. 2007;97(5):993-1000. doi: 10.1017/S0007114507669189 [DOI] [PubMed] [Google Scholar]
30.Jurj AL, Wen W, Xiang YB, et al. Reproducibility and validity of the Shanghai Men’s Health Study physical activity questionnaire. Am J Epidemiol. 2007;165(10):1124-1133. doi: 10.1093/aje/kwk119 [DOI] [PubMed] [Google Scholar]
31.Matthews CE, Shu XO, Yang G, et al. Reproducibility and validity of the Shanghai Women’s Health Study physical activity questionnaire. Am J Epidemiol. 2003;158(11):1114-1122. doi: 10.1093/aje/kwg255 [DOI] [PubMed] [Google Scholar]
32.Rao C, Yang G, Hu J, Ma J, Xia W, Lopez AD. Validation of cause-of-death statistics in urban China. Int J Epidemiol. 2007;36(3):642-651. doi: 10.1093/ije/dym003 [DOI] [PubMed] [Google Scholar]
33.Sudlow C, Gallacher J, Allen N, et al. UK biobank: an open access resource for identifying the causes of a wide range of complex diseases of middle and old age. PLoS Med. 2015;12(3):e1001779. doi: 10.1371/journal.pmed.1001779 [DOI] [PMC free article] [PubMed] [Google Scholar]
34.Bradbury KE, Young HJ, Guo W, Key TJ. Dietary assessment in UK Biobank: an evaluation of the performance of the touchscreen dietary questionnaire. J Nutr Sci. 2018;7:e6. doi: 10.1017/jns.2017.66 [DOI] [PMC free article] [PubMed] [Google Scholar]
35.Conroy MC, Lacey B, Bešević J, et al. UK Biobank: a globally important resource for cancer research. Br J Cancer. 2023;128(4):519-527. doi: 10.1038/s41416-022-02053-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
36.International Agency for Research on Cancer, World Health Organization . Tobacco smoking. IARC Monographs on the Evaluation of the Carcinogenic Risk of Chemicals to Humans, volume 38. 1986. Accessed July 14, 2026. https://publications.iarc.who.int/56 [PubMed]
37.International Agency for Research on Cancer, World Health Organization . Alcohol drinking. IARC Monographs on the Evaluation of Carcinogenic Risks to Humans, volume 44. 1988. Accessed July 14, 2026. https://publications.iarc.who.int/62
38.Cogliano VJ, Baan R, Straif K, et al. Preventable exposures associated with human cancers. J Natl Cancer Inst. 2011;103(24):1827-1839. doi: 10.1093/jnci/djr483 [DOI] [PMC free article] [PubMed] [Google Scholar]
39.World Cancer Research Fund/American Institute for Cancer Research . Diet, nutrition, physical activity and cancer: a global perspective. Continuous update project expert report 2018. Accessed September 2, 2026. dietandcancerreport.org
40.Doll R, Payne P, Waterhouse JA. Cancer Incidence in Five Continents: A Technical Report. Vol 3. Springer; 1966. doi: 10.1007/978-3-642-85849-9 [DOI] [Google Scholar]
41.Fay MP, Feuer EJ. Confidence intervals for directly standardized rates: a method based on the gamma distribution. Stat Med. 1997;16(7):791-801. doi: 10.1002/(SICI)1097-0258(19970415)16:7<791::AID-SIM500>3.0.CO;2-# [DOI] [PubMed] [Google Scholar]
42.Tiwari RC, Clegg LX, Zou Z. Efficient interval estimation for age-adjusted cancer rates. Stat Methods Med Res. 2006;15(6):547-569. doi: 10.1177/0962280206070621 [DOI] [PubMed] [Google Scholar]
43.Jensen OM, Parkin DM, MacLennan R, Muir CS, Skeet RG, eds. Cancer Registration: Principles and Methods. IARC Scientific Publication No. 95. International Agency for Research on Cancer; 1991
44.Sjölander A, Vansteelandt S. Doubly robust estimation of attributable fractions in survival analysis. Stat Methods Med Res. 2017;26(2):948-969. doi: 10.1177/0962280214564003 [DOI] [PubMed] [Google Scholar]
45.Chen L, Lin DY, Zeng D. Attributable fraction functions for censored event times. Biometrika. 2010;97(3):713-726. doi: 10.1093/biomet/asq023 [DOI] [PMC free article] [PubMed] [Google Scholar]
46.Kitagawa EM. Components of a difference between two rates. J Am Stat Assoc. 1955;50(272):1168-1194. [Google Scholar]
47.GBD 2021 Causes of Death Collaborators. Global burden of 288 causes of death and life expectancy decomposition in 204 countries and territories and 811 subnational locations, 1990-2021: a systematic analysis for the Global Burden of Disease Study 2021. Lancet. 2024;403(10440):2100-2132. doi: 10.1016/S0140-6736(24)00367-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
48.Crabb DW, Matsumoto M, Chang D, You M. Overview of the role of alcohol dehydrogenase and aldehyde dehydrogenase and their variants in the genesis of alcohol-related pathology. Proc Nutr Soc. 2004;63(1):49-63. doi: 10.1079/PNS2003327 [DOI] [PubMed] [Google Scholar]
49.Wilsnack RW, Vogeltanz ND, Wilsnack SC, et al. Gender differences in alcohol consumption and adverse drinking consequences: cross-cultural patterns. Addiction. 2000;95(2):251-265. doi: 10.1046/j.1360-0443.2000.95225112.x [DOI] [PubMed] [Google Scholar]
50.Fitzgerald N, Angus K, Emslie C, Shipton D, Bauld L. Gender differences in the impact of population-level alcohol policy interventions: evidence synthesis of systematic reviews. Addiction. 2016;111(10):1735-1747. doi: 10.1111/add.13452 [DOI] [PubMed] [Google Scholar]
51.Renehan AG, Zwahlen M, Egger M. Adiposity and cancer risk: new mechanistic insights from epidemiology. Nat Rev Cancer. 2015;15(8):484-498. doi: 10.1038/nrc3967 [DOI] [PubMed] [Google Scholar]
52.Wulan SN, Westerterp KR, Plasqui G. Ethnic differences in body composition and the associated metabolic profile: a comparative study between Asians and Caucasians. Maturitas. 2010;65(4):315-319. doi: 10.1016/j.maturitas.2009.12.012 [DOI] [PubMed] [Google Scholar]
53.Murphy N, Jenab M, Gunter MJ. Adiposity and gastrointestinal cancers: epidemiology, mechanisms and future directions. Nat Rev Gastroenterol Hepatol. 2018;15(11):659-670. doi: 10.1038/s41575-018-0038-1 [DOI] [PubMed] [Google Scholar]
54.Lin Y, Zheng L, Fang K, Zheng Y, Wu J, Zheng M. Proportion of liver cancer cases and deaths attributable to potentially modifiable risk factors in China. Int J Epidemiol. 2023;52(6):1805-1814. doi: 10.1093/ije/dyad100 [DOI] [PubMed] [Google Scholar]
55.Huang DQ, El-Serag HB, Loomba R. Global epidemiology of NAFLD-related HCC: trends, predictions, risk factors and prevention. Nat Rev Gastroenterol Hepatol. 2021;18(4):223-238. doi: 10.1038/s41575-020-00381-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
56.Liu Z, Jiang Y, Yuan H, et al. The trends in incidence of primary liver cancer caused by specific etiologies: Results from the Global Burden of Disease Study 2016 and implications for liver cancer prevention. J Hepatol. 2019;70(4):674-683. doi: 10.1016/j.jhep.2018.12.001 [DOI] [PubMed] [Google Scholar]
57.Swanson JM. The UK Biobank and selection bias. Lancet. 2012;380(9837):110. doi: 10.1016/S0140-6736(12)61179-9 [DOI] [PubMed] [Google Scholar]

Associated Data

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

Supplementary Materials