Adiposity, Adulthood Weight Change, and Risk of Incident Hepatocellular Carcinoma

With an annual global incidence of over 500,000 cases per year, hepatocellular carcinoma (HCC) represents the fifth most common malignancy and the third-leading cause of cancer-related mortality, worldwide (1). In the United States, the incidence of HCC has risen sharply in the past 20 years, in parallel with the epidemic of obesity (2, 3). Despite the growing burden of HCC, it carries a grim prognosis, with limited treatment options and an estimated 5-year survival of 18% (4). Consequently, an urgent need remains to develop effective strategies to prevent the development of HCC.

In the United States, the average adult gains between 0.5 and 1.0 kg per year, beginning in early adulthood (5). Such slow weight gain may go unnoticed for many years, until it ultimately results in overweight and obesity (5). Adulthood weight gain has been associated with adverse health outcomes, including type 2 diabetes, coronary heart disease, and colorectal cancer (6, 7). However, to date, most epidemiologic studies of obesity and HCC incidence have focused on attained adiposity [i.e., body weight, body mass index (BMI), waist circumference, etc.] assessed once, at a single point in time (8–10). Among the few prospective studies to assess dynamic changes in adiposity during adulthood in relation to HCC risk, all have lacked prospectively updated data during adulthood (11, 12), or focused exclusively on selected subpopulations, such as persons with intentional weight loss (13), without comprehensively evaluating measures of both total-body adiposity (i.e., weight, BMI) and central adiposity (i.e., waist circumference). Consequently, the impact of changes in adulthood weight, BMI, or waist circumference on the development of incident HCC is unclear. Moreover, whether the effects of increasing weight or BMI on HCC risk might vary according to their timing in adulthood is undefined.

It is plausible that maintaining a stable weight and normal BMI and waist circumference during adulthood—by preventing long-term weight gain—could represent an important strategy for HCC primary prevention. Such an approach would be welcome for its practicality, because achieving and sustaining weight loss is typically far more difficult after a person has already developed obesity (14). Thus, we leveraged two nationwide, prospective cohort studies to comprehensively evaluate the impact of adulthood BMI, waist circumference, and adulthood weight gain on the subsequent risk of developing incident HCC.

The Nurses' Health Study (NHS) prospectively enrolled 121,700 female registered nurses, ages 30–55 years, beginning in 1976, and the Health Professionals Follow-up Study (HPFS) prospectively enrolled 51,529 male health professionals, ages 40–75 years, in 1986 (15, 16). Since enrollment, participants have returned biennial questionnaires, providing prospectively updated data on lifestyle, medical history, physical activity, and disease outcomes, with follow-up that consistently exceeds 90% (17). For the current study, we included all individuals who reported early adulthood weight (i.e., age 18 years among women; age 21 years among men). We excluded adults with missing data on weight at baseline (1980 in NHS, 1986 in HPFS), or a prior diagnosis of cancer (except nonmelanoma skin cancer), consistent with prior studies (18, 19). The remaining 125,264 participants (77,238 women, 48,026 men) were eligible for analysis.

The NHS and HPFS cohorts were approved by the Institutional Review Boards (IRB) of the Brigham and Women's Hospital (Boston, MA) and the Harvard T.H. Chan School of Public Health (Boston, MA), and those of participating registries as required (IRB approval ID: 2009P001907). This study was conducted in accordance with the U.S. Common Rule; informed consent was implied by the return of self-administered questionnaires.

In the NHS cohort, recalled weight at age 18 years was queried in 1980, and in the HPFS cohort, weight at age 21 years was queried in 1986. In prior validation studies, the correlation coefficients between recalled and measured weights in these cohorts were 0.87 in women and 0.80 in men (20, 21). Beginning with the enrollment questionnaire, participants reported their height and current weight; on each subsequent biennial questionnaire, participants have updated their current weight. From these data, BMI was updated prospectively, every 2 years. In a prior validation study, the Pearson correlation coefficient between self-reported current weight and technician-measured weights was 0.97 for both women and men (22). Waist circumference was first reported in the 1986 questionnaire in NHS, and in a supplementary 1987 mailing in HPFS, in a subset of participants (22). Self-measured waist circumference in both cohorts has been validated against two technician-measured waist circumference measurements in a sample of 123 men and 140 women, respectively, with correlation coefficients of 0.89 (NHS) and 0.95 (HPFS; ref. 22).

Overall adulthood weight change was defined as the difference in weight (kilograms) between early adulthood (i.e., age 18 years among women; age 21 years among men) and present, using prospectively updated weights over each 2-year questionnaire cycle (7). In additional analyses, we also examined weight change during specific periods of adulthood, including in early adulthood (i.e., between age 18 years or 21 years and study baseline), and in mid-life [i.e., between baseline (in 1986) and present]. For each weight change exposure, we used the distribution of weight change in each cohort to classify participants in five categories (6, 7, 23): (i) weight loss ≥−5 kg, (ii) stable weight, defined as either weight loss between 0 and <−5 kg, or weight gain between 0 and <5 kg (reference category), (iii) moderate weight gain between ≥5 and 10 kg, (iv) substantial weight gain between >10.0 and 19.9 kg, and (v) extreme weight gain of ≥20.0 kg. We chose 10 kg as the cutoff for substantial weight gain based on prior studies (6, 7), and based on the median (10.0 kg in women, 7.8 kg in men) and mean (12.2 kg in women; 9.0 kg in men) values. The cutoff for the extreme weight gain group (20 kg) was based on selection of the highest 20% of participants (6, 23).

In the NHS and HPFS, self-reported diagnoses of liver cancer were obtained from biennial questionnaires, and all participants reporting primary liver cancer were asked for permission to acquire and review their medical records, relevant imaging and histopathology. A study physician, blinded to exposure information, reviewed all records to confirm HCC diagnoses, and extracted information from the medical charts regarding anatomic features and the presence of underlying cirrhosis and/or viral hepatitis from each HCC case.

Detailed and validated data regarding age, race/ethnicity, smoking status, alcohol consumption, medication use, diet, physical activity, and personal medical history were collected at baseline and updated every 2–4 years. For details regarding covariate ascertainment and definitions, please see the Supplementary Materials and Methods (Supplementary Appendix). Briefly, our main multivariable model included the following a priori confounders: age, race/ethnicity, smoking status, alcohol consumption (g/day), physical activity [metabolic equivalent task (MET)-hours/week], regular use of aspirin, a healthy dietary pattern [continuous Alternative Healthy Eating Index 2010 (AHEI-2010) score], and comorbid type 2 diabetes, dyslipidemia and hypertension. In additional analyses, we constructed separate models further adjusting for self-reported viral hepatitis and cirrhosis, consistent with prior work (24). With the exception of race/ethnicity, all covariates were prospectively updated over each 2-year questionnaire cycle, to better address misclassification or measurement error over long-term follow-up.

For analyses of waist circumference, we excluded any participant lacking initial waist circumference measurements in NHS (in 1986; n = 33,065 excluded) and HPFS (in 1987; n = 21,014 excluded), and for these analyses, follow-up began in 1986 (NHS) and 1987 (HPFS) when the information on waist circumference was first collected.

Three covariates had missing data: physical activity, dietary quality, and smoking (maximum percent missing = 3% for all). For missing data on physical activity and dietary quality, data were replaced using median values; for missing data on smoking status, participants were assigned to the never-smoker category, as in prior work (6).

Person-time accrued from the date of return of the initial questionnaire until the first recorded date of death, HCC diagnosis or end of follow-up (January 31, 2012 in HPFS; June 1, 2012 in NHS). First, we examined HCC risk according to attained adiposity at specific timepoints in adulthood, including young-adult BMI [i.e., age 18 years (women), or age 21 years (men)], mid-life BMI (i.e., at baseline), updated BMI (i.e., throughout adulthood), and updated waist circumference. For all models, we tested for linear trend using continuous variables.

Next, we analyzed adulthood weight change, and specifically evaluated (i) overall adulthood weight change, (ii) early adulthood weight change, and (iii) mid-life weight change, as defined previously. P for linear trend was calculated per each 1-kg increase among those with stable weight or weight gain. We also constructed separate models additionally adjusting for young-adult BMI, mid-life BMI, updated BMI or waist circumference.

We used Cox proportional hazards modeling stratified by age and calendar year, to calculate multivariable-adjusted HRs and 95% confidence intervals (CI), while accounting for a priori covariates (defined above and in the Supplementary Materials and Methods). For analyses of young-adult BMI and mid-life BMI, covariates were modeled at baseline; for analyses of updated BMI, waist circumference and weight change, all covariates (with the exception of race/ethnicity) were updated biennially and modeled as time-varying covariates. The proportionality assumption was not violated. All analyses were conducted separately in each cohort, and then pooled using an inverse variance-weighted fixed-effects meta-analysis.

To determine whether attained adulthood BMI confers additional HCC risk beyond weight change, we conducted nested analyses, comparing models that included only continuous adulthood weight change, to models that included both continuous adulthood weight change and continuous BMI in young adulthood, in mid-life, or updated throughout adulthood.

We conducted several sensitivity analyses to test the robustness of our results. First, we repeated the primary analysis after excluding any HCC case diagnosed within the first 4 years of follow-up, or within the first 8 years of follow-up, to address potential reverse causation (18, 19). Second, we applied an alternative definition of stable adulthood weight (i.e., weight change of −2 to <2 kg, rather than weight change of −5 to <5 kg). Third, because underlying cirrhosis can impact body weight and increase HCC risk, we excluded any participant with established cirrhosis at baseline; we also constructed separate models further adjusting for incident viral hepatitis or cirrhosis, reported during each biennial follow-up (24). Finally, we compared our results using a complete case analysis, in which we excluded any person with any missing baseline covariate data.

All statistical analyses were performed using SAS version 9.4 (SAS Institute), with a two-sided significance P < 0.05.

This study demonstrates in two nationwide, prospective cohorts that dynamic weight gain during adulthood is a significant, independent risk factor for the development of incident HCC. Even among adults with a normal body mass index, significant excess risk of developing HCC was found in those with higher waist circumference, or weight gain during adulthood of ≥10 kg or more. Together, these findings underscore the importance of maintaining a stable adulthood weight, for the primary prevention of HCC.

The funding sources did not participate in the design or conduct of this study; collection, management, analysis or interpretation of the data; or preparation, review or approval of the article.

No additional data are available.

The senior author (X. Zhang) affirms that the article is an honest, accurate, and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned (and if relevant, registered) have been explained.

Baseline characteristics of study participants across categories of adulthood weight change are shown in Table 1. During adulthood, women gained a median of 10.0 kg [interquartile range (IQR), 3.6–19.1 kg], and men gained a median of 7.8 kg (IQR, 2.3–15.6 kg). Overall, men and women with more adulthood weight gain were more likely to have type 2 diabetes and hypertension and to be physically inactive.

Over 3,676,549 person-years of follow-up, we documented 158 incident HCC cases (98 cases among women; 60 cases among men). An elevated mid-life BMI and an elevated updated BMI were both significantly associated with increased HCC risk, after multivariable adjustment (P_trend = 0.034 and 0.004, respectively; Table 2), including after further adjusting for young-adult BMI (P_trend = 0.022 and 0.001, respectively). Specifically, after adjusting for young-adult BMI, each additional unit of higher mid-life BMI contributed to a 5% higher multivariable-adjusted risk of HCC (aHR = 1.05; 95% CI, 1.02–1.07), while each additional unit of updated BMI contributed to an 8% higher multivariable-adjusted risk of HCC (aHR = 1.08; 95% CI, 1.06–1.10). These associations remained significant in separate analyses of women and men.

Young-adult BMI was significantly and positively associated with increased HCC risk in the combined cohort (aHR for young-adult BMI < 19 vs. ≥25 kg/m² = 2.12, 95% CI = 1.02–4.42; aHR per each 1 unit of higher young-adult BMI = 1.05, 95% CI = 1.02–1.09; P_trend = 0.019; Table 2). However, this association was markedly attenuated after further accounting for mid-life BMI (corresponding aHR for young-adult BMI = 1.02; 95% CI, 0.98–1.05; P_trend = 0.14), or updated BMI (corresponding aHR = 1.02; 95% CI, 0.97–1.06; P_trend = 0.39).

In further analyses, we evaluated waist circumference among the 92,199 individuals with waist circumference data beginning in 1986 (NHS) and 1987 (HPFS). Supplementary Table S1 outlines the characteristics of this subgroup according to adulthood weight change categories. An elevated waist circumference was significantly associated with increased HCC risk, after multivariable adjustment [aHR per each 1-cm increase in waist circumference, 1.04 (95% CI = 1.02–1.11), P_trend = 0.016; Supplementary Table S2), including among subgroups of adults with BMI < 25 kg/m² [corresponding aHR = 1.02 (95% CI = 1.00–1.06), P_trend = 0.038] and among those with overweight/obesity [corresponding aHR = 1.03 (95% CI = 1.00–1.10), P_trend = 0.015; Supplementary Table S3].

HCC risk increased with overall adulthood weight gain (aHR per each 1-kg gain = 1.03; 95% CI = 1.01–1.08; P_trend = 0.010), including after further adjusting for young-adult BMI (corresponding aHR = 1.03; 95% CI = 1.01–1.09; P_trend = 0.010; Table 3). Specifically, compared with participants with stable weight during adulthood, those who gained between 5–<10 kg, 10–<20 kg, and ≥20 kg had multivariable aHRs of, 1.40 (95% CI, 0.67–2.16), 2.09 (95% CI, 1.11–3.95), and 2.61 (95%CI, 1.42–5.22), respectively. These associations persisted after further adjusting for mid-life BMI, updated BMI or continuous waist circumference (P_trend = 0.018, 0.039, and 0.012, respectively). In nested models, the association between overall adulthood weight change and incident HCC risk was substantially modified by the addition of updated BMI to the multivariable model (log-likelihood P = 0.048). However, it was not modified by young-adult BMI, mid-life BMI or waist circumference (log-likelihood P = 0.54, 0.13, and 0.38, respectively). Overall adulthood weight loss was not significantly associated with HCC incidence.

We also conducted analyses focused specifically on early-adulthood weight gain (i.e., between age 18 or 21 years and baseline) and on mid-life weight gain (i.e., between the baseline date and present). Early-adulthood weight gain was significantly and positively associated with increased HCC risk (aHR per each 1-kg gain = 1.03; 95% CI = 1.01–1.07; P_trend = 0.001), even after accounting for young-adult BMI (corresponding aHR = 1.03; 95% CI = 1.01–1.09; P_trend < 0.001; Supplementary Table S4). When we focused on mid-life weight gain, this gradient of increasing risk was modestly attenuated, but remained statistically significant (aHR per each 1-kg gain = 1.03; 95% CI = 1.01–1.08; P_trend = 0.012), even after further adjusting for baseline BMI (corresponding aHR = 1.02; 95% CI = 1.00–1.09; P_trend = 0.016; Supplementary Table S5).

Our findings were robust across all sensitivity analyses, including after using an alternative definition of stable adulthood weight (Supplementary Table S6). Results for weight change also persisted after we excluded all incident HCC cases diagnosed within 4 or within 8 years (Supplementary Table S7), or anyone with established cirrhosis at baseline (Supplementary Table S8). Our findings were not materially altered after further adjustment for incident viral hepatitis and cirrhosis during follow-up (Supplementary Table S9). Finally, when we conducted a complete case analysis restricted to participants with complete covariate data at baseline (n = 121,845), our findings again were consistent (Supplementary Table S10).

Within two nationwide, prospective cohorts, we investigated dynamic measures of adiposity throughout adulthood in relation to HCC incidence. HCC risk was significantly higher in participants with an elevated young-adult BMI, mid-life BMI, updated BMI and with progressive weight gain during overall adulthood. The significant, positive associations between updated BMI, adulthood weight gain and HCC risk also persisted regardless of early-adult BMI. Overall, the lowest HCC risk was observed in adults who maintained a stable weight during adulthood. In contrast, the highest HCC risk was found in participants with the most weight gain during adulthood, particularly among those with the highest BMI or waist circumference. Collectively, these findings indicate that both an elevated BMI and progressive adulthood weight gain are significant and independent risk factors for the development of HCC. Thus, our data suggest that maintaining a stable weight during adulthood, specifically by preventing weight gain, could represent an important public health strategy for HCC prevention.

To date, the impact of dynamic changes in adiposity during adulthood on HCC risk is undefined. Although several prior studies have demonstrated a link between obesity and HCC incidence, most have focused on attained adiposity, measured at single points in time (8–10, 25). Among the few published studies to evaluate longitudinal changes in body weight or BMI, all have lacked prospectively updated assessments of adiposity (11, 12), or specific estimates of HCC risk (26), or focused on subgroups, such as patients with intentional weight loss (13). Thus, by leveraging two nationwide populations with detailed and prospectively updated data regarding body weight and adiposity over 26 years of follow-up, the current study can provide a more comprehensive examination of adiposity in relation to HCC risk. Besides confirming the association between attained adiposity and HCC (8–10), our study additionally demonstrates that adulthood weight gain is a significant and independent HCC risk factor, and further that this risk is augmented in individuals with the highest attained BMI or waist circumference.

Within the general population, the adverse effects of excess adiposity and weight gain are well described (6, 7). More recently, clinical studies have demonstrated that interventions to maintain a healthy body weight can improve insulin sensitivity, normalize inflammatory markers, and reduce liver fat and inflammation, all of which are potential contributors to obesity-related hepatocarcinogenesis (27–29). Although it remains unknown whether such short-term benefits might translate to reduced HCC incidence, prior studies focused on other health outcomes—including other cancers—have demonstrated substantial benefits from maintaining a healthy body weight throughout adulthood (6).

Recent evidence suggests that the influence of adiposity and weight gain on carcinogenesis varies across the life course. Early-life adiposity may contribute to an increased risk of incident endometrial, esophageal, and colorectal cancer (30–32), but a lower risk of breast cancer (33), while later-adulthood obesity has been associated with an increased risk of colorectal cancer among men (31, 34). Our findings suggest that weight gain in early-adulthood and in mid-life both contribute to increased HCC risk, and the magnitude of that risk appears to be strongest in those who begin gaining excess weight in early adulthood, which ultimately results in obesity during later adulthood. Collectively, this suggests that the long-term, sustained effects of increasing adiposity may confer excess HCC risk, a finding that is consistent with our current understanding of the prolonged latency of HCC. Indeed, the progressive accumulation of adiposity has been shown to promote hyperinsulinemia and lipotoxicity, perturbations of the gut microbiome and stimulation of the IGF1 signaling pathway, which over many years may promote hepatocarcinogenesis (35–38). Nevertheless, further investigation will be needed to fully characterize the precise mechanisms that underpin the critical period during which excess adiposity is most strongly associated with HCC risk.

This study is strengthened by a prospective design, relatively large sample size, and the inclusion of nationwide populations of men and women with detailed, well-validated assessments of body weight (20), adiposity, lifestyle, and potential confounding factors. Using repeated assessments reduces measurement errors, while simultaneously addressing real-world changes in weight throughout adulthood. We also accounted for wide ranges of time between exposure assessment and the development of incident HCC, which minimizes the possibility of bias from reverse causation. In support of this, when the primary analysis was repeated among persons without established cirrhosis, the positive association between weight gain and HCC risk was modestly strengthened.

We acknowledge several limitations. First, early adulthood weight was recalled, and there remains the potential for systematic error in the assessment of weight change. However, prior validation studies in these cohorts demonstrate the validity of recalled weights (correlation coefficients between recalled and measured weight, 0.87 in women and 0.80 in men; refs. 20, 21). Moreover, because weight data were collected long before the development of HCC, minimizing the potential for reverse causation. Second, we acknowledge that our participants are predominantly white, and that one cohort included only women, while the second included only men, which could impact generalizability. Furthermore, in the overall study population, cirrhosis and hepatitis B or C infection status were self-reported, as was alcohol consumption, and we also lacked details regarding liver fibrosis stages or adherence to HCC screening; collectively, these factors highlight the need for future research in larger and more diverse, prospective populations with both men and women, that include more detailed data regarding underlying liver disease. However, our age- and sex-specific rates of incident HCC, and the distributions of our well-validated covariate data (15, 16), closely approximate other population-based cohorts, and the homogeneity of health care access and socioeconomic status in these cohorts minimizes confounding and enhances internal validity. In addition, the magnitude of early adulthood weight gain in our cohorts is comparable with that found in the general U.S. population (i.e., adulthood weight gain over comparable periods of time in the National Health and Nutrition Examination Survey was approximately 10 kg; ref. 39), which further underscores the generalizability of our results. Third, we acknowledge that we had limited statistical power in some of our subgroup and sensitivity analyses. We also did not include data regarding the speed and trajectories of weight changes during adulthood, which requires further investigation. Finally, although reverse causation is possible, the exposures were assessed many years prior to study outcomes, and our findings were similar even after excluding HCC cases diagnosed within 8 years of follow-up.

In two prospective cohorts of U.S. men and women, both excess adiposity and adulthood weight gain were significantly and independently associated with increased risk of developing incident HCC. The lowest HCC risk was observed among adults who maintained a stable weight during adulthood. In contrast, increasing adulthood weight gain contributed to significant, positive gradient of increasing HCC risk, across all levels of BMI or waist circumference. Together, these findings support public health efforts designed to maintain a healthy body weight by preventing adulthood weight gain, for the long-term prevention of HCC.

T.G. Simon reports grants from Amgen and other support from Aetion during the conduct of the study. C.S. Fuchs reports personal fees from Amylin Pharma, AstraZeneca, Bain Capital, CytomX Therapeutics, Daiichi-Sankyo, Eli Lilly, Entrinsic Health, Evolveimmune Therapeutics, Genentech, Merck, Taiho, and Unum Therapeutics outside the submitted work; and C.S. Fuchs also serves as a Director for CytomX Therapeutics and owns unexercised stock options for CytomX and Entrinsic Health. C.S. Fuchs is a co-Founder of Evolveimmune Therapeutics and has equity in this private company. C.S. Fuchs had provided expert testimony for Amylin Pharmaceuticals and Eli Lilly. C.S. Fuchs is now an employee of Genentech and Roche. A.T. Chan reports personal fees from Bayer Pharma AG, Pfizer Inc., and Boehringer Ingelheim, and grants from Zoe Global Ltd. outside the submitted work. No disclosures were reported by the other authors.

T.G. Simon: Conceptualization, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. M.N. Kim: Data curation, validation, visualization, writing–review and editing. X. Luo: Visualization, writing–review and editing. X. Liu: Visualization, writing–review and editing. W. Yang: Data curation, visualization, writing–review and editing. Y. Ma: Data curation, formal analysis, validation, visualization, writing–review and editing. D.Q. Chong: Validation, writing–review and editing. C.S. Fuchs: Visualization, methodology, writing–review and editing. M. Stampfer: Conceptualization, validation, visualization, methodology, writing–review and editing. E.L. Giovannucci: Conceptualization, data curation, validation, visualization, methodology, writing–review and editing. A.T. Chan: Conceptualization, resources, supervision, validation, visualization, methodology, writing–review and editing. X. Zhang: Conceptualization, resources, data curation, formal analysis, supervision, validation, investigation, visualization, methodology, writing–review and editing.

We would like to thank the participants and staff of the NHS and the HPFS for their valuable contributions as well as the following state cancer registries for their help: AL, AZ, AR, CA, CO, CT, DE, FL, GA, ID, IL, IN, IA, KY, LA, ME, MD, MA, MI, NE, NH, NJ, NY, NC, ND, OH, OK, OR, PA, RI, SC, TN, TX, VA, WA, WY. The authors assume full responsibility for analyses and interpretation of these data.

This work is supported by UM1 CA186107 (NHS infrastructure grant), P01 CA87969 (NHS program grant for cancer research), U01 CA167552 (HPFS infrastructure grant), NIH K24 DK098311 (A.T. Chan), NIH R21CA238651 (X. Zhang), American Cancer Society Research Scholar Grant (RSG NEC-130476, X. Zhang), NIH K23 DK122104 (T.G. Simon). Simon is supported by the Harvard University Center for AIDS Research (CFAR) and by the Dana Farber/Harvard Cancer Center (DFHCC) GI SPORE. X. Zhang is supported by the Zhu Family Center for Global Cancer Prevention at Harvard T. H. Chan School of Public Health and the Dana Farber/Harvard Cancer Center (DFHCC) GI SPORE.

A.T. Chan is a Stuart and Suzanne Steele MGH Research Scholar.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.