A validated estimate of visceral adipose tissue volume in relation to cancer risk

Abstract

Background: Despite the recognized role of visceral adipose tissue in carcinogenesis, its independent association with cancer risk beyond traditional obesity measures remains unknown because of limited availability of imaging data.

Methods: We developed an estimation equation for visceral adipose tissue volume using elastic net regression based on demographic and anthropometric data in a subcohort of participants in the UK Biobank (UKB; n = 23 148) with abdominal magnetic resonance imaging scans. This equation was externally validated in 2713 participants from the 2017-2018 National Health and Nutrition Examination Survey according to sex, age, and race groups. We then applied the equation to the overall UKB cohort of 461 665 participants to evaluate the prospective association between estimated visceral adipose tissue and cancer risk using Cox proportional hazards models. We also calculated the population attributable risk of cancer associated with estimated visceral adipose tissue and body mass index (BMI).

Results: Estimated visceral adipose tissue showed a high correlation with measured visceral adipose tissue in internal and external validations (r = 0.81-0.86). During a median 12-year follow-up in the UKB, we documented 37 397 incident cancer cases; estimated visceral adipose tissue was statistically significantly associated with elevated risk of obesity-related and individual cancers, independent of BMI and waist circumference. Population attributable risk for total cancer associated with high (quartiles 2-4 vs 1) estimated visceral adipose tissue (9.0% for men, 11.6% for women) was higher than high BMI (quartiles 2-4 vs 1 = 5.0% for men, 8.2% for women).

Conclusions: Estimated visceral adipose tissue showed robust performance in UKB and National Health and Nutrition Examination Survey and was associated with cancer risk independent of BMI and waist circumference. This study provides a potential clinical tool for visceral adipose tissue estimation and underscores that visceral adipose tissue can be an important target for cancer prevention.

Figure 1.

The validation of the estimation equation for visceral adipose tissue in the test set of the imaging subcohort of the UK Biobank and in the National Health and Nutrition Examination Survey (NHANES) cohort. A) and (B) The distribution of the measured and the estimated visceral adipose tissue in the test set of the imaging subcohort for men (n = 3389) (A) and women (n = 3609) (B). The paired t test was conducted to test the statistical difference between the 2 variables. C) and (D) The Spearman correlation between the measured and estimated visceral adipose tissue in the test set of the imaging subcohort for men (C) and women (D). E) and (F) The Spearman correlation accounting for complex survey between the measured visceral adipose tissue volume by dual-energy x-ray absorptiometry and the estimated visceral adipose tissue volume in the NHANES cohort for men (n = 1378) (E) and women (n = 1335) (F). VAT = visceral adipose tissue.

Figure 2.

Multivariable-adjusted hazard ratios for the associations of estimated visceral adipose tissue, body mass index (BMI), and waist circumference at the initial assessment visit (2006-2010) with subsequent cancer risk in men. A) Shows the marginal and mutually adjusted associations of estimated visceral adipose tissue, BMI, and waist circumference with cancer risk per 90th-to-10th percentile difference increment (5.0 L for estimated visceral adipose tissue, 9.7 kg/m² for BMI, and 27 cm for waist circumference). The asterisk “*” indicates the association with a false discovery rate less than 0.05. B) Summarizes the changes in marginal vs mutually adjusted associations of estimated visceral adipose tissue, BMI, and waist circumference with risk of cancers that showed a marginal association (total, obesity-related, gastrointestinal [GI], colorectal, kidney, esophageal, pancreatic, stomach, and liver cancers). The multivariable-adjusted hazard ratios were adjusted for age at baseline, education level, race, height, total physical activity, smoking status and pack-years of smoking, alcohol use, regular use of aspirin, consumption of processed meat, vegetable intake, fresh fruit intake, cereal intake, circulating vitamin D levels, family history of cancer, and history of bowel cancer screening. Total cancer included all cancers, except for nonmelanoma skin cancer and nonfatal prostate cancer. Obesity-related cancer was composed of esophageal cancer, stomach cancer, liver cancer, kidney cancer, pancreatic cancer, colorectal cancer, gallbladder cancer, postmenopausal breast cancer, ovarian cancer, thyroid cancer, malignant meninges, and multiple myeloma. GI cancer included malignant neoplasm of the esophagus, stomach, small intestine, colon, rectosigmoid junction, rectum, anus and anal canal, liver and intrahepatic bile ducts, gallbladder, other and unspecified parts of biliary tract, pancreas, and other and ill-defined digestive organs. CI = confidence interval; eVAT = estimated visceral adipose tissue; HR = hazard ratio; WC = waist circumference.

Figure 3.

Multivariable-adjusted hazard ratios for the associations of estimated visceral adipose tissue, body mass index (BMI), and waist circumference at the initial assessment visit (2006-2010) with subsequent cancer risk in women. A) Shows the marginal and mutually adjusted associations of estimated visceral adipose tissue, BMI, and waist circumference with cancer risk per 90th-to-10th percentile difference increment (3.3 L for estimated visceral adipose tissue, 11.9 kg/m² for BMI, and 30 cm for waist circumference). The asterisk “*” indicates the association with a false discovery rate less than 0.05. B) Summarizes the changes in marginal vs mutually adjusted associations of estimated visceral adipose tissue, BMI, and waist circumference with risk of cancers that showed a marginal association (total, obesity-related, gastrointestinal [GI], postmenopausal breast, endometrial, pancreatic, kidney, and bladder cancers). Multivariable-adjusted hazard ratios were adjusted for age at baseline, education level, race, height, total physical activity, smoking status and pack years of smoking, alcohol use, regular use of aspirin, consumption of processed meat, vegetable intake, fresh fruit intake, cereal intake, circulating vitamin D levels, family history of cancer, and history of bowel cancer screening. Total cancer included all cancers, except for nonmelanoma skin cancer and nonfatal prostate cancer. Obesity-related cancer was composed of esophageal cancer, stomach cancer, liver cancer, kidney cancer, pancreatic cancer, colorectal cancer, gallbladder cancer, postmenopausal breast cancer, ovarian cancer, thyroid cancer, malignant meninges, and multiple myeloma. GI cancer included malignant neoplasm of the esophagus, stomach, small intestine, colon, rectosigmoid junction, rectum, anus and anal canal, liver and intrahepatic bile ducts, gallbladder, other and unspecified parts of biliary tract, pancreas, and other and ill-defined digestive organs. CI = confidence interval; eVAT = estimated visceral adipose tissue; HR = hazard ratio; WC = waist circumference.

Figure 4.

Population attributable risk (PAR)% of cancer prevented had the higher quartiles (Q2-4) of estimated visceral adipose tissue or body mass index (BMI) been reduced to quartile 1 in the overall UK Biobank cohort in men (A) and women (B), with the distribution of other risk factors unchanged. The quartiles of the estimated visceral adipose tissue in men were quartile 1 (≥0.81, <3.76 L), quartile 2 (≥3.76, <5.01 L), quartile 3 (≥5.01, <6.38 L), and quartile 4 (≥6.38 L); the quartiles of estimated visceral adipose tissue in women were quartile 1 (≥0.43, <1.73 L), quartile 2 (≥1.73, <2.53 L), quartile 3 (≥2.53, <3.50 L), and quartile 4 (≥3.50 L). The quartiles of BMI in men were quartile 1 (≥18.51, <25.06 kg/m²), quartile 2 (≥25.06, <27.32 kg/m²), quartile 3 (≥27.32, <30.00 kg/m²), and quartile 4 (≥30.00 kg/m²); the quartiles of BMI in women were quartile 1 (≥18.50, <23.55 kg/m²), quartile 2 (≥23.55, <26.15 kg/m²), quartile 3 (≥26.15, <29.66 kg/m²), and quartile 4 (≥29.66 kg/m²). Other risk factors fixed in the PAR% calculation included age (high [50 years and older] vs low [younger than 50 years]), race (White vs non-White), family history of cancer (yes vs no), ever having bowel cancer screening (no vs yes), college degree (no vs yes), regular aspirin consumption (yes vs no), intake of processed meat (any intake vs never), smoking (any smoking vs never smoking), alcohol drinking (any drinking vs never drinking), physical activity (<28.5 vs ≥28.5 metabolic equivalents hours per week), intake of vegetables (<4 vs ≥4 heaped tablespoons per day), intake of fresh fruit (<2 vs ≥2 pieces per day), intake of cereal (<5 vs ≥5 bowls per week), serum vitamin D level (<50 vs ≥50 nmol/L), height (> vs ≤ 162 cm for women, > vs ≤ 176 cm for men). Only positive PARs are shown in the figure. Full PAR results for all cancers are presented in Supplementary Table 6 (available online). For liver cancer, the PAR% represented the cases prevented had estimated visceral adipose tissue or BMI reduced from quartile 4 to quartile 3-1 level, because the PAR for reducing from quartile 4-2 to quartile 1 level could not be calculated because of the negative point estimates of hazard ratio for quartile 3 and quartile 2 (vs reference level quartile 1). Total cancer included all cancers, except for nonmelanoma skin cancer and nonfatal prostate cancer. Obesity-related cancer was composed of esophageal cancer, stomach cancer, liver cancer, kidney cancer, pancreatic cancer, colorectal cancer, gallbladder cancer, postmenopausal breast cancer, ovarian cancer, thyroid cancer, malignant meninges, and multiple myeloma. GI cancer included malignant neoplasm of the esophagus, stomach, small intestine, colon, rectosigmoid junction, rectum, anus and anal canal, liver and intrahepatic bile ducts, gallbladder, other and unspecified parts of biliary tract, pancreas, and other and ill-defined digestive organs. eVAT = estimated visceral adipose tissue.