Few studies have examined the relationship of lifestyle factors with mortality among patients with colorectal cancer.
Few studies have examined the relationship of lifestyle factors with mortality among patients with colorectal cancer.
Among NIH-AARP Diet and Health study participants, 4213 colon and 1514 rectal cancer cases were identified through linkage to state cancer registries and determined date and cause of death using the National Death Index. Lifestyle factors were assessed at baseline and included: healthy diet (measured by Healthy Eating Index 2005 [HEI-2005]), body mass index (BMI), physical activity, alcohol consumption and smoking. The association of factors was examined individually and combined into a lifestyle score with 5-year mortality from all-causes, colorectal cancer, and cardiovascular disease (CVD). Relative risks (RRs) and 95% confidence intervals (CIs) were estimated using Cox proportional hazards models.
Among colon cancer survivors, smokers had increased risk of total mortality (RR = 1.74; 95% CI = 1.45-2.08) and colorectal cancer mortality (RR = 1.46; 95% CI = 1.17-1.82), compared to never smokers. Obese (BMI, ≥ 30) individuals had increased risk of all death (RR = 1.19; 95% CI = 1.02-1.39) and CVD death (RR = 1.84; 95% CI = 1.05-3.23), compared to normal weight (BMI, 18.5 to < 25) individuals. Compared to those with the lowest lifestyle score, those with the highest score had a 34% lower risk of all-cause mortality (RR = 0.66; 95% CI = 0.50-0.87). Among rectal cancer survivors, individuals in the highest quintile of HEI-2005 scores had reduced all-cause mortality (RR = 0.60; 95% CI = 0.42-0.86) compared to those in the lowest. Higher combined lifestyle scores were associated with a 46% lower risk of total mortality (0.54; 0.32-0.91).
Healthier lifestyle before cancer diagnosis was associated with improved overall survival after diagnosis with colorectal cancer. Cancer 2014;120:1540–1547 2014.
The number of colorectal cancer survivors in the United States is estimated at more than 1 million. The Institute of Medicine reports numerous gaps in the clinical care of survivors, among them being the need for guidance on “maintaining a healthy lifestyle and preventing recurrent or new cancers.” However, whereas there is abundant evidence linking diet and lifestyle to colorectal cancer incidence, evidence for their effect on patient survival after a cancer diagnosis is only now emerging.
Accumulating evidence from observational studies suggests prediagnosis and postdiagnosis body weight,[6-11] and prediagnosis and postdiagnosis physical activity,[12-15] may have an impact on colorectal cancer survival. Fewer studies have examined the effect of postdiagnosis diet quality, or prediagnosis alcohol consumption, or smoking.[17, 18] Although evidence is limited, the American Cancer Society (ACS) has published Guidelines on Nutrition and Physical Activity for Cancer Survivors, which can be summarized as follows: 1) achieve and maintain a healthy weight throughout life; 2) be physically active; and 3) eat a healthy diet, with an emphasis on plant foods. The World Cancer Research Fund and American Institute for Cancer Research Expert Report made similar recommendations that also include limiting alcohol consumption.
Because cancer survivors are at increased risk of second cancers and other comorbid conditions, following an overall healthy lifestyle pattern may be important for improving survival outcomes. However, no studies have evaluated the combined effect of healthy behaviors on cancer survival. Therefore, we examined the relationship of prediagnosis lifestyle factors, alone and in combination, on 5-year all-cause, colorectal cancer–specific, and cardiovascular (CVD) mortality among colon and rectal cancer cases in the National Institutes of Health (NIH)-AARP Diet and Healthy Study cohort.
Details of the study methods and population have been previously described. Briefly, the NIH-AARP Diet and Health Study participants consisted of AARP members aged 50 to 71 years, residing in 6 states (California, Florida, Pennsylvania, New Jersey, North Carolina, and Louisiana) and 2 metropolitan areas (Atlanta, Ga, and Detroit, Mich). Participants completed mailed questionnaires at baseline in 1995-1996. The study was approved by the Special Studies Institutional Review Board of the US National Cancer Institute (NCI).
Cancer cases were ascertained through December 31, 2006, by linkage to state cancer registries in the study area, and areas where participants tended to move during follow-up (Arizona, Nevada, and Texas). More than 90% of cancer cases were identified through this linkage. Registry data include diagnosis date, morphology, stage, grade, and first course of treatment.
We identified 7107 first primary invasive colorectal cancer cases among the 492,182 men and women who satisfactorily completed the baseline questionnaire and had no history of cancer (except for nonmelanoma skin cancer), end-stage renal disease, or self-reported poor health at baseline. We further excluded those: with a colorectal cancer diagnosis <1 year after baseline (n = 793); who reported total energy intakes greater than 2 times the interquartile range (n = 59); who had missing height or weight or had body mass index (BMI) < 18.5 kg/m2 (n = 220); had missing physical activity (n = 67); or unknown smoking history (n = 241). We included 4213 colon and 1514 rectal cancer cases in our analysis.
Mortality and cause of death were ascertained by linkage with the Social Security Administration Death Master File and the National Death Index Plus through December 31, 2008. All-cause mortality was defined as death from any cause whereas colorectal cancer-specific mortality was defined as having colorectal cancer as the underlying cause of death. CVD death was defined as: disease of the heart (International Classification of Diseases, 10th Revision [ICD-10] codes I00-I09, I11, I13, I20-I51), hypertension (I10, I12), cerebrovascular diseases (I60-I69), atherosclerosis (I70), or aortic aneurism (I71).
Diet (including alcohol consumption) was assessed with a 124-item food frequency questionnaire, in which participants reported the frequency of food and beverage consumption for the previous 12 months. The instrument used 10 predefined categories of intake ranging from “never” to “6+ times per day” for beverages, and “never” to “2+ times per day” for solid foods. Portion size and nutrient intakes were calculated using the 1994 to 1996 US Department of Agriculture's Continuing Survey of Food Intakes by Individuals databases.
Dietary quality was determined by applying the Healthy Eating Index 2005 (HEI-2005), which assesses conformance to the 2005 Dietary Guidelines for Americans. The HEI-2005 is energy adjusted using a density method, and scores range from zero to 100 points, with higher scores representing better conformity to dietary guidelines. Participants reported body weight, height, physical activity, and smoking status in the baseline questionnaire. BMI was computed as weight in kilograms divided by the square of height in meters. Physical activity questions queried the frequency that participants engaged in ≥20 minutes activity that resulted in increased breathing, heart rate, or perspiration. Smoking questions included baseline smoking status and time since quitting for former smokers.
Colon and rectal cancer cases were analyzed separately. Lifestyle factors were examined individually as follows: HEI-2005 scores were categorized into quintiles and alcohol intake was divided into 3 categories: nondrinking, moderate drinking (≤2 drinks/day for men, or ≤1 drink/day for women), and heavier drinking (>2 drinks/day for men or >1 drink/day for women), with one drink equal to ∼13 g alcohol. BMI was grouped into 3 levels: normal weight (18.5 to <25 kg/m2), overweight (25 to <30 kg/m2), and obese (≥30 kg/m2). Because underweight individuals may have unique underlying health problems that may affect survival, they were excluded. Physical activity was categorized into 5 levels (never or rarely, 1-3 times/month, 1-2 times/week, 3-5 times/week, and ≥5 times/week) and smoking history was divided into 4 categories (never smoked, quit ≥10 years ago, quit 1-9 years ago, quit <1 year ago, or current smoker).
To compute the composite lifestyle score, we dichotomized each lifestyle variable and assigned one point for meeting the recommendation and zero points for not meeting it. Participants received one point for each of the following: BMI from 18.5 to < 25 kg/m2, ≥3 episodes/week physical activity, HEI score within the top 2 quintiles, consuming ≤1 drink/day (women) or 2 drinks/day (men), and not smoking (including former smokers who had quit at least 1 year before baseline). We then summed the points for all variables to produce a single score with values ranging from 0 (worst score) to 5 (best score).
Relative risks (RRs) and 95% confidence intervals (CIs) were estimated using Cox proportional hazards models, with age as the underlying time metric. Because mortality follow-up began at the age of cancer diagnosis, participants had different lengths of time between baseline measurements and cancer diagnosis (“lag time”). Exploratory analyses revealed interactions between lag time and the composite lifestyle score, but limiting follow-up time to 5 years after diagnosis eliminated this interaction. Participants' follow-up ended at either age at censoring or age at death. Censoring events were reaching 5 years follow-up after diagnosis, or at study end (December 31, 2008), whichever was first. Models were adjusted for sex, cancer stage, first course of treatment, and family history of colon cancer. We further adjusted multivariate models for lag time differences in tertiles (1 year to <4.3 years, 4.3 years to <7.3 years, ≥7.3 years). In additional analyses, we stratified Cox models by cancer stage, sex, smoking, and BMI.
Among 4213 colon cancer cases, 1273 cases died during 5-year follow-up (856 due to colorectal cancer, 125 to other cancers, 108 to CVD, and 184 deaths from other causes). Among 1514 rectal cancer cases, 454 were deceased by 5 years (301 colorectal cancer, 49 other cancer, 43 CVD, and 61 deaths from other causes).
Table 1 presents the distribution of demographic, tumor, treatment, and lifestyle variables across the range of lifestyle scores. Among both colon and rectal cancer cases, those with higher scores tended to be older at diagnosis, women, and more highly educated.
|Colon Cancer Cases (N = 4213)||Rectal Cancer Cases (N = 1514)|
|Mean age at diagnosis, y||68.4||69.0||69.6||70.3||70.5||68.4||68.3||68.9||69.1||69.8|
|College graduate, %||33.4||32.9||33.0||42.4||46.8||29.9||29.0||30.7||34.9||46.5|
|Family history of colon cancer, %||10.0||10.1||11.2||9.7||13.7||5.4||6.8||7.4||7.3||9.3|
|II (regional by extension)||15.6||12.9||10.9||10.9||8.5||10.9||6.8||9.5||9.8||8.1|
|III (regional with lymph node)||17.1||16.8||180.0||18.6||21.8||10.9||17.7||17.1||15.0||10.5|
|Mean Body Mass Index (kg/m2)||28.8||29.3||28.3||26.3||22.9||28.1||29.1||27.5||25.7||23.4|
|Mean Healthy Eating Index 2005 score||52.4||58.4||66.9||73.5||77.5||53.2||57.8||65.4||72.6||76.5|
|Physically active, %a||4.6||16.0||43.7||80.4||100.0||5.4||20.8||49.3||76.2||100.0|
|Moderate drinker, %b||27.8||78.3||92.0||97.4||100.0||22.8||74.7||88.4||97.3||100.0|
Table 2 presents the association between lifestyle variables and 5-year mortality among colon cancer cases. Prediagnosis smoking and obesity were associated with an increased risk of 5-year all-cause mortality. Compared to never-smokers, those who were current smokers had 74% increased risk of death from any cause within 5 years, and had 46% increased risk of death from colorectal cancer. Compared to normal weight individuals, those who were obese had about a 19% increased risk of death from any cause, and approximately 84% increased risk of CVD death. There were also significant trends of decreased CVD mortality with increasing HEI-2005 scores and with alcohol consumption.
|All-Cause Mortality||Colorectal Cancer Mortality||CVD Mortality|
|Deaths (n)||RRa||95% CI||Deaths (n)||RRa||95% CI||Deaths (n)||RRa||95% CI|
|Healthy Eating Index-2005|
|Quintile 1 (21.8-56.7)||268||1.00||170||1.00||32||1.00|
|Quintile 2 (56.8-65.2)||264||1.07||(0.89-1.28)||177||1.15||(0.92-1.44)||18||0.51||(0.28-0.92)|
|Quintile 3 (65.3-71.4)||239||0.99||(0.82-1.19)||168||1.09||(0.87-1.38)||15||0.42||(0.22-0.81)|
|Quintile 4 (71.5-76.9)||265||1.08||(0.89-1.31)||179||1.16||(0.91-1.47)||26||0.64||(0.35-1.15)|
|Quintile 5 (77.0-90.2)||237||0.95||(0.78-1.16)||162||0.99||(0.77-1.27)||17||0.45||(0.23-0.87)|
|Body mass index (kg/m2)|
|Normal weight (BMI 18.5-24.9) 2524.9)||376||1.00||267||1.00||22||1.00|
|Overweight (BMI 25-29.9)||543||1.02||(0.88-1.17)||357||0.97||(0.82-1.15)||53||1.72||(1.02-2.88)|
|Never or rarely||278||1.00||179||1.00||23||1.00|
|5 or more times/week||229||1.04||(0.87-1.25)||168||1.16||(0.93-1.45)||17||0.90||(0.47-1.72)|
|Quit 10 or more years ago||484||1.04||(0.90-1.20)||329||0.95||(0.80-1.12)||44||1.21||(0.74-1.99)|
|Quit 9 years ago or less||207||1.29||(1.09-1.54)||118||1.01||(0.81-1.27)||23||1.88||(1.06-3.48)|
The associations between individual lifestyle factors and mortality among rectal cancer cases are presented in Table 3. Compared to nonsmokers, those who smoked had 34% increased risk of all-cause mortality, and those with highest HEI-2005 scores had 40% decreased risk of 5-year all-cause mortality, compared to those with the lowest scores. There was a significant trend of decreased risk of colorectal cancer specific mortality with increasing HEI-2005 scores. Prediagnosis physical activity and alcohol intake were not associated with either all-cause mortality or colorectal cancer–specific mortality among colon or rectal cancer cases.
|All-Cause Mortality||Colorectal Cancer Mortality||CVD Mortality|
|Deaths (n)||RRb||95% CI||Deaths (n)||RRb||95% CI||Deaths (n)||RRb||95% CI|
|Healthy Eating Index-2005|
|Quintile 1 (21.8-56.7)||122||1.00||83||1.00||8||1.00|
|Quintile 2 (56.8-65.2)||102||0.87||(0.66-1.15)||71||0.89||(0.64-1.26)||11||1.37||(0.53-3.54)|
|Quintile 3 (65.3-71.4)||98||0.96||(0.71-1.30)||64||0.94||(0.65-1.37)||10||0.92||(0.32-2.64)|
|Quintile 4 (71.5-76.9)||75||0.93||(0.67-1.29)||42||0.81||(0.53-1.23)||12||1.67||(0.62-4.45)|
|Quintile 5 (77.0-90.2)||57||0.60||(0.42-0.86)||41||0.64||(0.41-0.99)||2||0.28||(0.06-1.43)|
|Body mass index (kg/m2)|
|Normal weight (BMI 18.5-24.9)||147||1.00||92||1.00||16||1.00|
|Overweight (BMI 25-29.9)||198||0.85||(0.68-1.07)||135||0.92||(0.70-1.22)||17||0.78||(0.38-1.61)|
|Never or rarely||92||1.00||59||1.00||8||1.00|
|1-3 times per month||72||1.08||(0.78-1.48)||53||1.23||(0.84-1.80)||6||0.98||(0.31-3.07)|
|1-2 times per week||99||0.87||(0.65-1.16)||70||0.92||(0.64-1.31)||10||1.08||(0.41-2.87)|
|3-4 times per week||112||1.01||(0.76-1.35)||69||0.95||(0.66-1.37)||12||1.17||(0.46-3.00)|
|5 or more times per week||79||0.84||(0.61-1.15)||50||0.78||(0.53-1.15)||7||0.95||(0.33-2.77)|
|Quit 10 or more years ago||195||1.23||(0.96-1.58)||130||1.28||(0.94-1.75)||15||1.01||(0.41-2.47)|
|Quit 9 years ago or less||82||1.38||(1.02-1.87)||53||1.29||(0.89-1.87)||11||2.74||(1.07-7.03)|
Higher composite lifestyle scores were associated with reduced all-cause mortality (Fig. 1) among colon and rectal cancer cases. Compared to those who met 1 or fewer recommendations, colon cancer cases who met all 5 had a 34% reduced risk of death (RR = 0.66; 95% CI = 0.50-0.87; P-for-trend = .006; Fig. 1A), and rectal cancer cases had a 46% reduced risk of death (RR = 0.54; 95% CI = 0.32-0.91; P-for-trend = .01, Fig. 1B). Higher lifestyle scores were also associated with reduced risk of colorectal cancer mortality among rectal cancer cases (Fig. 1C), but not among colon cancer cases (Fig. 1D). There were no significant interactions between the composite lifestyle score and sex, smoking, BMI, or cancer stage.
In this study, we found that a combined lifestyle score was associated with reduced all-cause mortality among both colon and rectal cancer cases. Several modifiable prediagnostic lifestyle factors contributed to this association. In colon cancer cases, obesity and smoking were related to an increased risk of all-cause mortality. Smoking was also associated with increased risk of colorectal cancer–specific mortality whereas the HEI-2005, obesity, and alcohol consumption were related to the risk of death due to CVD. In rectal cancer cases, higher HEI-2005 scores were associated with decreased risk of all-cause and colorectal cancer mortality, and smoking was related to an increased risk of all-cause and CVD mortality.
Given the well-known negative impact of smoking on health, it is not surprising that prediagnosis smoking is associated with increased risk of death among colorectal cancer cases. Similar to our finding, a study of incident colorectal cancer cases identified in the Surveillance, Epidemiology, and End Results (SEER) registry in Washington State found that prediagnosis smoking was associated with increased risk of all-cause mortality (RR = 1.51; 95% CI = 1.24-1.83) and colorectal cancer–specific mortality among colorectal cancer cases (RR = 1.30; 95% CI = 7.09-1.74). In another study of stage III colon cancer cases, prediagnosis smoking was also associated with increased risk of death from any cause (RR = 1.65; 95% CI = 1.12-2.42).
In our study, prediagnosis obesity was associated with poorer overall survival among colon cancer cases. Other studies have also found that prediagnosis BMI, or body weight, body fat percentage, and waist circumference were associated with poorer overall and disease-specific survival among colorectal cancer cases. However, studies of postdiagnosis BMI and mortality in colorectal cancer cases have reported inconsistent results.[7-11]
Pre- or postdiagnosis physical activity has been found to be associated with lower risk of mortality among colorectal cancer cases in several studies.[5, 12-15] The lack of association between prediagnosis physical activity and mortality in our study could be partly due to differences in measurement instruments, in the time between measurement and diagnosis, or differences in adjustment factors.
In the only other study that examined dietary patterns and survival among colon cancer patients, a postdiagnosis Western dietary pattern (characterized by high intakes of meat, fat, and desserts) was associated with poorer survival. Among rectal cancer cases, we found lower all-cause mortality among those with higher scores on the HEI-2005, whereas we observed no association among colon cancer cases. This lack of association might be explained by the fact that the HEI-2005 is constructed for overall health rather than cancer prevention or survival. For example, the HEI-2005 awards points for meat consumption without separating red meat (a known colorectal cancer risk factor) from other, leaner sources of animal protein. However, this index incorporates several other features that may partially mitigate these limitations. Specifically, HEI-2005 rewards points for lower consumption of foods thought to contribute to poor health outcomes, such as sugar and saturated fats (ie, calories from Solid Fats, Alcoholic beverages, and Added Sugars [SoFAAS]). Thus, points awarded for consumption of red meat would be somewhat ameliorated by a low score in the SoFAAS category. In addition, the HEI-2005 awards points for consumption of whole fruit and whole grains, which is consistent with ACS and American Institute for Cancer Research guidelines for cancer prevention and cancer survivors.[4, 21] Furthermore, this index has already been shown to be related to colorectal cancer risk in this study population, and in contrast to the data-driven approach of the study by Meyerhardt et al, has the ability to be directly compared to other studies that use the same index.
Only one other study has examined alcohol in relationship to survival among colorectal cancer cases, and it found no association of alcohol intake with all-cause or disease-specific mortality. We found that alcohol intake was related to lower risk of CVD mortality, but not to colorectal cancer mortality among colon cancer cases. More research is needed on dietary patterns and alcohol in relation to cancer survival.
We found significantly reduced risk of 5-year all-cause mortality among colon and rectal cancer cases with a higher combined lifestyle score. Energy balance factors such as diet, body weight, and physical activity influence colorectal cancer risk and may also impact cancer recurrence and progression. Proposed mechanisms for the association between energy balance and cancer risk or progression include peptide hormones (ie, leptin and adiponectin), insulin and insulin-like growth factor 1 signaling, and inflammation. Furthermore, BMI, diet, and physical activity have been shown to be associated with genetic and epigenetic changes in colon and rectal tumors, and these changes could have implications for recurrence, progression, and therefore survival.
Our study has several limitations. One important limitation is that lifestyle information was collected only once, before cancer diagnosis, and we do not know how lifestyle changes after diagnosis might affect survival outcomes among colorectal cancer cases. Clearly, maintaining healthy behaviors in adulthood is consistent with current disease prevention guidelines, and with recommendations for the prevention of comorbidities and improved quality of life among cancer survivors. However, currently there is insufficient evidence to recommend lifestyle changes to cancer survivors specifically for the purpose of improving survival outcomes. At most, these results suggest a possible impact of lifestyle behaviors on cancer survival, and support the need for further prospective research on this topic.
This analysis was also limited by the fact that subjects had different lengths of time between the lifestyle assessment and cancer diagnosis (lag time). We therefore adjusted the analyses for lag time and limited follow-up time to 5 years. Also, whereas we had cancer registry treatment data on first course of therapy within 12 months of cancer diagnosis, we did not have extensive clinical data, including information on recurrence or other nonmortality outcomes. Small sample sizes may have limited our ability to detect associations, and small cell sizes in some analyses could produce spurious associations. In addition, because true associations of lifestyle factors and cancer survival outcomes are not known, we weighted each recommended lifestyle factor equally when creating a single lifestyle score. Equal weighting may give undue emphasis to weak risk factors and underestimate the effects of others. Therefore, the protective trend associated with higher lifestyle scores should be emphasized rather than the exact magnitude of effect for the combined score. Furthermore, we cannot rule out confounding of health behaviors with unmeasured factors such as socioeconomic status and health care access. Strengths of our study include the large sample size that allowed us to examine colon and rectal cancer cases separately, and the collection of multiple health behaviors, which enabled the examination of a lifestyle score computed based on current recommendations for cancer survivors.
In summary, we found that healthier lifestyle before colorectal cancer diagnosis was associated with improved overall survival after diagnosis with colorectal cancer. Our study adds to a growing body of literature demonstrating the importance of maintaining a healthy lifestyle before cancer diagnosis. More research is needed to investigate the effect of these lifestyle factors on health in colorectal cancer survivors.
Funding provided by the Intramural Research Program of the National Institutes of Health, National Cancer Institute.
The authors made no disclosures.