Adult Overweight and Survival from Breast and Colorectal Cancer in Swedish Women

The estimated global prevalence of overweight and obesity has increased from 25% in 1980 to 39% in 2014. Over the same period, the prevalence of obesity has doubled, from 6% to 13% (1). In many countries, these changes have impacted the occurrence of, and mortality from, main noncommunicable diseases including cancer (2, 3). In 2012 alone, half a million new cancer cases globally could be attributed to high body mass index (BMI); in Sweden, high BMI was responsible for 4.5% of all cancers (2,247 cases) occurring in that year (4). Furthermore, as the obesity epidemic progresses and life expectancy continues to rise, more individuals experience longer durations of being overweight during their life course, with still largely unknown consequences on the population burden of cancer.

Yet, most studies to-date have investigated the BMI–cancer association using exposure information on BMI measured at a single point in time and typically obtained at study baseline, meaning that insights into the cumulative impact of overweight and obesity during the life course remain scarce. This particularly concerns the concept of overweight intensity, which combines the degree and duration of overweight over time and can be considered the obesity-equivalent of pack-years of cigarette smoking. From a biological perspective, early life exposure and accumulated exposure to overweight increase the risk and severity of insulin resistance, chronic inflammation, oxidative DNA damage, and alterations in endogenous hormone metabolism—all key mechanisms on the obesity–cancer pathway (5–7). Furthermore, obesity at different time points throughout the life course has been associated with different types of and kinetics for cancer development (8, 9). Although previous studies have confirmed the impact of age-dependent and cumulative effects of weight change on cancer occurrence (10, 11), their impact on prognosis after diagnosis is still poorly understood (12). A meta-analysis of prospective cohort studies concluded that overweight and obesity, assessed prediagnostically at one point in time, are associated with poorer overall and breast cancer survival (13), and that the evolution of body size from childhood to adulthood has a long-term influence on breast cancer survival (14).

In view of the ongoing and increasing obesity epidemic worldwide, insights into the relation between cumulative overweight exposure and cancer survival have become vital for the planning and implementation of effective treatment and follow-up strategies. In this study, we aim to assess the association between excess body weight during adulthood and overall survival in postmenopausal patients with breast and colorectal cancer in a cohort of Swedish women.

The Swedish Women's Lifestyle and Health (WLH) study is a prospective cohort study that started in 1991 with the aim to investigate the association between lifestyle factors and cancer in young women. Details of the study design have been described previously (15). In brief, inclusion criteria were: ages between 29 and 49 years at baseline, and resident in the Uppsala Health Care Region. In total, 96,000 women were randomly selected from the Swedish Population Registry and sent baseline questionnaires, which were completed by 49,259 (51%) women. In 2003, this final cohort of women received a follow-up questionnaire, with a response rate of 73%. Self-reported lifestyle and anthropometric measures (body height and weight) were included in both the baseline and the follow-up questionnaire. Women included in the final cohort were followed for a mean duration of 20.7 years. The study was approved by the regional Ethical Committee at Uppsala University, and the Ethical Committee at Karolinska Institutet, Stockholm. For this current analysis, women with a cancer diagnosis before age 50 (N = 1,311) and with either unreliable BMI information (height lower than 130 cm or BMI lower than 16 kg/m²; N = 17) or BMI information dating exclusively less than 1 year before cancer diagnosis (N = 879) were excluded, leaving a study population of 47,051 women for analysis.

Information on vital status, inpatient care, emigration, and incident cancer cases was obtained from linkage to Swedish registries through the individually unique Swedish national registration number, allowing for virtually complete follow-up. Cases of incident breast and colorectal cancer were identified through linkage with the Swedish Cancer Registry (ICD7 codes 170, 153–154).

Based on the population of women who contributed at least one BMI measurement after the age of 20 and before the end of the study, death, or a diagnosis of cancer, whichever came first, we modeled the trajectory of BMI as a function of age using a linear mixed effect model with a quadratic effect of time and with random intercepts and slopes assumed to be normally distributed. We used information on weight and height from all 47,051 included women to construct this model in order to achieve a robust approximation of the relationship between age and BMI.

For each woman, the model-based BMI trajectory was thus described by a quadratic polynomial from which several variables were created: (i) the mean BMI during early adulthood, computed as the integral under the BMI trajectory divided by 30 (that is, the length in years of the period defining early adulthood); (ii) the time spent with a BMI exceeding 25 (overweight + obesity) during early adulthood; and (iii) the predicted slope (i.e., yearly BMI increase) between ages 20 and 50. This exposure window was selected to allow every study participant to, in principle, accumulate the same overweight duration (maximum 30 years) and separate this period from the window of risk of death (starting at age 50).

The study populations for this part of the analysis were constituted of 1,241 women who were diagnosed with primary breast cancer and 259 women who were diagnosed with primary colorectal cancer after the age of 50. Person-years of follow-up were calculated from the date of cancer diagnosis to the date of death, the date of last contact, or the end of the study follow-up (December 31, 2012), whichever came first. Analyses were based on Cox proportional regression models. The following variables were used as adjusting factors in the analysis of the association between death in patients with breast and colorectal cancer and BMI-related variables: age at diagnosis (continuous), smoking (never vs. ever), alcohol consumption (continuous), and past history of diabetes mellitus (only for breast cancer). For the BMI-related variables, we used the following coding: mean BMI during early adulthood was categorized according to WHO cut-offs into less than 22 kg/m², between 22 and 25 kg/m² and more than 25 kg/m²; time spent with BMI over 25 kg/m² was categorized into never (reference), less than 12.1 years, and more than 12.1 years with a maximum of 30 years, where the values defining the categories were based on the median of the distribution of non-nil times spent with overweight or obesity. The predicted slope of the BMI trajectory (i.e., the yearly BMI increase) between ages 20 and 50 was modelled continuously per SD increment in SD.

A secondary analysis was restricted to 830 women diagnosed with breast cancer (respectively, 167 women diagnosed with colorectal cancer) for whom information on tumor–node–metastasis (TNM) staging was available. Although we present here the results using the fully-adjusted model, results of univariate analyses are presented in Supplementary Tables S1 and S2.

All analyses were performed with the R statistical software (version 3.4.0; R Development Core Team, 2017).

Cases with a mean BMI above the overweight threshold (≥25 kg/m²) during adulthood (age 20–50) were slightly older at cancer diagnosis and more likely to have had a history of diabetes or high blood pressure (at baseline) when compared with cases with a mean BMI below 25 kg/m² (Table 1). Relative to patients with breast and colorectal cancer with a mean BMI below 25 kg/m², the distribution of stage of disease at diagnosis was slightly poorer among those with elevated BMI levels, in particular for colorectal cancer (13.7% vs. 27.5% stage IV).

Overall, survival in postmenopausal patients with breast and colorectal cancer worsened with an increasing mean BMI during early adulthood (between age 20 and 50). After multivariate adjustment, HRs comparing a mean BMI ≥25 kg/m² to a mean BMI ≤22 kg/m² were 1.72 (95% CI, 1.05–2.83) and 2.41 (95% CI, 1.31–4.41), respectively, and linearly increased (Table 2; Fig. 1).

A longer duration of overweight and obesity (years with BMI ≥25 kg/m²) increased the risk of early death from postmenopausal breast cancer by 3% for every additional year spent with overweight in early adulthood (Table 2; Fig. 2). Yet, effects levelled off with increasing duration of overweight.

Similar associations were found for colorectal cancer, with a 2-fold risk of early death in patients who were overweight or obese for more than 12.1 years during adulthood and every additional year further increasing the risk by 4% (Table 2). Effects for both overweight duration and the mean yearly increase in BMI increased in a linear fashion (Figs. 2 and 3).

Repeating the analyses in a subset of cases with information available on stage of disease at diagnosis and additionally adjusting for this factor did not substantially change the results for postmenopausal breast cancer, yet the results for colorectal cancer became statistically insignificant (Supplementary Table S3). There was no evidence for a differential effect of the BMI-related variables by stage at diagnosis (Table 3; Supplementary Table S4), however risks were slightly higher at advanced disease stages (stages III–IV). Although associations tended to be stronger in nonsmoking women (Supplementary Tables S5 and S6), interaction terms for smoking and BMI-related variables were only statistically significant for the association between mean BMI and colorectal cancer.

A longer duration of overweight and obesity between ages 20 and 50 years increased the risk of dying in women with postmenopausal breast and colorectal cancer. Although this was largely independent of stage of disease at diagnosis, every additional year of life lived with a BMI above 25, as well as the intensity of overweight above this threshold, further increased the risk. Overweight intensity, which summarizes the degree of overweight over time, is therefore a useful measure of characterizing the obesity–cancer relationship that provides additional insights when compared with considering overweight at a single point in time. These findings point towards a long-lasting effect of overweight and obesity during young adulthood, impacting not only the risk to develop but also the chance of dying after a cancer diagnosis.

Although previous studies have explored the association between being overweight in adulthood and cancer incidence, only very little is known about its impact on survival after cancer diagnosis. Results from a French cohort study indicated that an increase in body size during adult life and a constantly large body size were associated with an increased risk of death from any cause in breast cancer survivors, independent of prediagnostic BMI (14). Other previous studies looked at weight gain and found that patients with breast cancer with the highest weight gain between ages 20 to 30 had an increased risk of death from any cause relative to women with a stable weight (16). Although the Nurses' Health Study concluded that prediagnostic weight was related to an elevated risk of death from breast cancer in nonsmoking women (17), other pooled and meta-analyses confirmed increased mortality risks irrespective of the smoking status and cause of death (18, 19). Our study is congruent with these findings and quantified the relationship in a more continuous fashion, showing that not only being overweight for longer periods during adulthood matters, but also the intensity of overweight (i.e., the combination of duration and degree of overweight). In addition, risks of dying related to ever being overweight were slightly higher among patients diagnosed with stage III or IV breast cancer, potentially pointing towards accelerated disease progression related to positive energy balance. High BMI may be detrimental for survival in patients with cancer via energy balance and inflammatory pathways in the tumor microenvironment and in the systemic circulation that stimulate tumor growth, progression, and metastasis (12).

Prediagnostic obesity and, to a lesser extent, overweight have also been linked to poorer outcomes in patients with colorectal cancer (20). Death rates from colon cancer increased with increasing BMI in a large cohort of American men and women (21) and also European cohorts (22). BMI was furthermore significantly associated with both disease-free and overall survival and constituted an independent prognostic determinant in clinical trial participants receiving adjuvant chemotherapy (23). Similarly, a recent U.S. study showed that obese patients with stage I to III disease had a 2-fold increased risk of death as compared with overweight patients (24). This was even more pronounced for patients who reported a stable weight (from pre- to peridiagnostic). The findings of our study add to this knowledge and provide additional insights into the dose–response relationship between prediagnostic adulthood BMI and survival from colorectal cancer.

Measuring lifetime exposures and relating these to cancer risk and survival has several methodologic challenges. In our study, prediagnostic BMI trajectories and overweight duration during adulthood (age 20–50) were estimated in order to allow every study participant to obtain the same exposure window that was subsequently linked to survival from postmenopausal cancer (defined as age >50 years). Using this approach, we separated periods of exposure and risk (of dying) and thereby circumvented possible bias related to overlaps between the 2 (e.g., someone with an event would by definition have a shorter exposure period). Yet, by taking this life course approach in assessing dose–response effects of overweight in relation to cancer outcomes, we were able to investigate the combination of the duration and the degree of overweight over time, which represents a major strength of our study and a rather novel approach. Capturing long-term exposure, this measure reflects underlying biological pathways with the potential for long-term overweight exposure to lead to chronic inflammation and altered hormone metabolism. The associations we found in this study were slightly stronger using overweight intensity when compared with mean BMI, suggesting that the former measure is more likely to appropriately capture the true overweight exposure.

Another main strength of this study is related to the cohort's size and prospective design that allowed for long-term follow-up of cancer incidence and deaths. Yet, to-date only relatively few deaths from breast and colorectal cancer occurred, potentially linked to the good and increasing prognosis of these sites, but also to the relatively young profile of the cohort (mean age at end of follow-up: 60.5 years). For this reason only overall survival could be studied, not allowing inferences as to whether the excess rate among overweight women is attributable to the well-established increase in baseline mortality or whether there is also a contribution from disease-specific mortality from breast and colorectal cancer, respectively. Also, predictors for overall mortality can differ from those of cancer-specific mortality. Furthermore, data on height and weight were self-reported using questionnaires at baseline and follow-up, with no information on their validity. Trajectories of BMI were thus modeled based on 2 time points only. As women typically tend to underestimate their weight, the results might be subject to exposure measurement error. With regard to BMI as a proxy for body fatness it should be noted that it is impossible to distinguish adiposity from muscularity using this measure. Women who were ever overweight might thus have a different body composition and fat distribution, and thereby, by definition, a different mortality risk. In this cohort of Swedish women, only a small proportion of women was ever obese (9.5%) during adulthood, meaning that an analysis for obesity was not possible as too few individuals/deaths contributed to this stratum. This could partly be explained by the so-called healthy volunteer bias. It was shown that in this cohort nonresponders were more likely to smoke, have a shorter education, and to be less physically active when compared with responders (15). In the absence of any information on the true age at menopause of the study participants, we defined age 50 as the cut-off point to distinguish pre- from postmenopausal cancers and to mark the upper limit of the exposure window for adulthood overweight ranging from age 20 to 50. Previous studies have suggested a mean age at menopause of 50.9 years in Swedish women (25). Only very limited data were available on clinical subtypes (e.g., hormone receptor status), treatment, and postdiagnostic weight. Weight changes after diagnosis can be greatly influenced by treatment (most notably chemotherapy; ref. 26), requiring sensible methodologic approaches to better understand the additional impact of treatment on weight trajectories covering pre-, peri-, and postdiagnostic periods in relation to prognosis. Also, the degree to which weight changes in childhood and adolescence influences adulthood BMI trajectories; this potentially contributed to the effects we found in this study and should be explored in the future. However the necessary data on childhood weight were not available in this study. Finally, some degree of residual confounding cannot be ruled out.

In conclusion, the findings of this study showed that adulthood overweight duration and intensity have a long lasting influence on survival after cancer diagnosis. Yet, the association remains complex and clinic-pathologic factors may explain parts of this relationship. Although these results need to be confirmed in larger studies and also in men (for colorectal cancer), this work highlights the need for effective prevention of overweight and obesity starting at an early age and over the life course, especially in high-risk populations. It also emphasizes the need to understand and improve long-term outcomes in patients with cancer with a history of obesity.

No potential conflicts of interest were disclosed.

Conception and design: M. Arnold, H. Charvat, H. Freisling, H.-O. Adami, I. Soerjomataram, E. Weiderpass

Development of methodology: M. Arnold, H. Charvat, H. Freisling, H.-O. Adami, I. Soerjomataram

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Arnold, H.-O. Adami, E. Weiderpass

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Arnold, H. Charvat, H. Freisling

Writing, review, and/or revision of the manuscript: M. Arnold, H. Charvat, H. Freisling, H. Noh, H.-O. Adami, I. Soerjomataram, E. Weiderpass

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H. Charvat, I. Soerjomataram

Study supervision: H.-O. Adami, I. Soerjomataram

The authors wish to thank all women who contributed to the study. This work was supported by the World Cancer Research Fund International (grant no. 2016/1636). Where authors are identified as personnel of the International Agency for Research on Cancer/World Health Organization, the authors alone are responsible for the views expressed in this article and they do not necessarily represent the decisions, policy, or views of the International Agency for Research on Cancer/World Health Organization.

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.