Lifestyle after Colorectal Cancer Diagnosis in Relation to Survival and Recurrence: A Review of the Literature

Two observational studies, the NHS I [23] and a German cohort of CRC survivors [24], assessed post-diagnosis dietary patterns in a population-based setting [23], while CALGB 89803 [42] reported results in the adjuvant setting (Table 1). Data-driven dietary patterns were assessed within NHS I [23] and CALGB 89803 [42]. Both studies observed patterns that were given the labels a “Western” and a “Prudent” dietary pattern. The Western dietary pattern was characterized by high- and low-fat dairy, refined grains, red and processed meats, desserts, and potatoes, while the Prudent dietary pattern was characterized by high intakes of fruits, vegetables, whole grains, and poultry.

For the Western dietary pattern, both studies reported an increased all-cause mortality risk [23, 42]. However, the association was statistically significant only in the adjuvant setting (CALBG: Q5 vs. Q1: HR 2.32; 95% CI 1.36–3.96; P trend <0.001) [42], and not in the population-based study (NHS I: Q5 vs. Q1: HR 1.32 (0.89–1.97); P trend = 0.23) [23]. Similarly, a statistically significant increased risk of colon cancer recurrence was reported in the adjuvant setting [42], while a non-significant positive association was reported for CRC mortality in the population-based study [23] (Table 1). For the Prudent dietary pattern, both studies reported statistically non-significant associations for all-cause mortality [23, 42], CRC-specific mortality [23], or colon cancer recurrence [42].

Furthermore, several a priori-defined dietary patterns were studied in the two population-based studies [23, 24] (Table 1). Of the a priori-defined dietary patterns, none has been studied in more than one cohort. Some a priori-defined dietary patterns were associated with lower risk of all-cause mortality, but not all [23, 24].

Only one small (n = 111) randomized dietary intervention trial among CRC survivors assessed associations with survival [78]. Throughout the 1.5 months of neoadjuvant radiotherapy patients with rectal cancer randomized to the intervention group received 6 weekly individualized nutrition counseling and education sessions using regular foods, while the control group maintained their usual diet. Overall, the main goal of the intervention was to enable every patient to achieve his or her calculated energy and protein requirements. After long-term follow-up (median follow-up 6.5 (range 4.9–8.1) years), CRC-specific survival was significantly longer in the intervention group after adjustment for age and disease stage (median survival 7.3 vs. 4.9 years).

Both NHS I [23] and CPS II Nutrition Cohort [40] reported on post-diagnosis red and processed meat intake, although the NHS I paper focused on dietary patterns (Table 1). The CPS II Nutrition Cohort also provided information regarding pre- to post-diagnosis change in red and processed meat consumption [40] (Table 2).

These two studies did not observe an association between red and processed meat intake and both all-cause mortality and CRC-specific mortality [23, 40]. Furthermore, changing meat intake from high (median or higher) before CRC diagnosis to low (below median) after diagnosis was not associated with lower mortality when compared to survivors with a consistently high intake [40].

Both the NHS I [23] and CALGB 89803 [43] reported on post-diagnosis sugar-sweetened beverage intake and CRC outcomes (Table 1).

Both studies [23, 43] reported increased all-cause mortality risk for sugar-sweetened beverage consumption after CRC diagnosis, of which the association in the NHS I was statistically significant [23]. Each additional serving of sugar-sweetened beverages (including fruit juices) after CRC diagnosis was associated with an 11% increased risk for all-cause mortality (HR 1.11; 95% CI 1.01–1.23) [23]. A similar relative risk was reported for CRC-specific mortality, although it was not statistically significant [23]. For colon cancer recurrence, CALGB 89803 reported a statistically significant increased recurrence risk for patients consuming ≥ 2 servings of sugar-sweetened beverages per day (HR 1.75; 95% CI 1.04–2.94) compared to those consuming < 2 servings per month (P trend = 0.04) [43].

Four population-based studies, NHS I [23, 44], HPFS [44], CPS II Nutrition cohort [27], and a Japanese cohort of CRC survivors [26], reported on post-diagnosis alcohol consumption and CRC outcomes (Table 1).

In NHS I, moderate drinking was used as the reference group and abstaining from alcohol consumption was associated with a statistically significant increased all-cause mortality risk (HR 1.30; 1.05–1.61) compared to women consuming 5–15 g of alcohol per day [23]. Drinking > 15 g/day (approximately 1.5 drinks) was not statistically significantly associated with increased mortality risk. Similarly, abstainers had a higher mortality risk than drinkers in the Japanese cohort [26] and after combining both NHS I and HPFS cohort data [44]. However, the CPS II Nutrition cohort reported that drinking alcohol after diagnosis was not associated with all-cause mortality [27]. For CRC-specific mortality, similar results were reported as for all-cause mortality (Table 1).

The CPS II Nutrition cohort also provided information regarding pre- to post-diagnosis change in alcohol consumption (Table 2). Participants who reported drinking before CRC diagnosis but stopped drinking alcohol after diagnosis had a statistically non-significant increased risk of all-cause and CRC-specific mortality compared to participants who continued to drink alcohol [27].

The intake of some foods, beverages, and nutrients were only reported in one study each (Table 1). Higher nut consumption was associated with lower risk of CRC mortality (HR/serving/day 0.69; 95% CI 0.49–0.97) in the NHS I, while no statistically significant association was reported for all-cause mortality [23]. Furthermore, no associations were observed within the NHS I with either all-cause mortality or CRC-specific mortality for vegetables, fruits, or whole grains [23]. However, in the Japanese study, lower green leafy vegetable intake after CRC diagnosis was associated with an increased all-cause mortality risk [26].

Higher milk intake was statistically significantly associated with lower all-cause mortality risk (Q4 vs. Q1: HR 0.72; 95% CI 0.55–0.94; P trend = 0.02) in the CPS II Nutrition Cohort [41]. A similar risk was reported for overall dairy consumption, although associations did not reach statistical significance [41]. Additionally, higher coffee intake was statistically significantly associated with lower all-cause mortality (≥ 4 vs. 0 cups/day: HR 0.66; 95% CI 0.37–1.18; P trend = 0.01) within CALGB 89803 [25]. No significant associations were reported for non-herbal tea intake [25].

Higher dietary glycemic load and total carbohydrate intake were statistically significant associated with an increased risk of mortality and recurrence in CALGB 89803 [45]. Higher total calcium intake was statistically significantly associated with both lower all-cause mortality and CRC-specific mortality in the CPS II Nutrition Cohort, while no significant associations were reported for vitamin D [41]. Also no significant associations were reported for intake of one-carbon nutrients (folate, vitamins B₆ and B₁₂) in NHS I [44].

One small randomized intervention trial which provided individualized nutritional counseling and education about regular foods suggest that making dietary changes may improve cancer-specific survival. No dietary pattern or food has been studied in more than two observational cohorts, with cancer recurrence only studied in one cohort in the adjuvant setting embedded in a randomized chemotherapy trial. While alcohol consumption has been studied more frequently, these studies often used abstainers as comparison group. Abstainers are probably an inappropriate reference group, as this group may, at least in part, include people who stopped drinking because of comorbidities or cancer-related symptoms. Overall, emerging evidence shows that diet after CRC diagnosis might affect survival, but further research is needed to clarify what aspects of diet are important and which dietary changes could affect survival.

For all-cause mortality, seven studies [26, 46‐52] were included in previous meta-analyses [7‐10]. These meta-analyses have found highest versus lowest post-diagnostic physical activity to be associated with 40% lower all-cause mortality risk [7‐10]. Five studies that were included in a dose-response meta-analysis showed a 28% lower risk of all-cause mortality (HR 0.72; 95% CI 0.65–0.80) for every 10 metabolic equivalent task-hour per week (MET-hours/week) increase in post-diagnosis physical activity [9], which is equivalent to current recommendations of 150 min/week of at least moderate intensity activity. For CRC-specific mortality, similar risk reductions were reported comparing high versus low physical activity after CRC diagnosis (HR 0.62; 95% CI 0.45–0.86) [11••] and for every 10 MET-hours/week increase in post-diagnosis physical activity (HR 0.75; 95% CI 0.65–0.85) [9].

The Australian cohort [48] and NHS I [46] also provided results on changes in physical activity and mortality outcomes in CRC patients (Table 2). An increase of physical activity > 2 h/week between 5 and 12 months post-diagnosis was statistically significantly associated with lower all-cause (HR 0.69; 95% CI 0.50–0.94) and CRC-specific mortality (HR 0.64; 95% CI 0.44–0.93) among Australian CRC survivors [48]. A pre- to post-diagnosis increase in physical activity showed a statistically significant lower all-cause and CRC-specific mortality risk in the NHS I [46], but no association was reported among Australian CRC survivors [48] (Table 2). The first randomized controlled trial designed primarily to assess the impact of physical activity on survival among colon cancer survivors is ongoing [83]. As of April 2017, the trial has enrolled 536 of its planned 972 participants [84] and only 1 year feasibility results have been published so far [85].

Three of the population-based studies, CPS II Nutrition Cohort [50], National Institutes of Health-AARP Diet and Health [51], and HPFS [53] also reported on post-diagnosis sedentary behavior and all-cause as well as CRC-specific mortality (Table 1). CPS II reported on leisure time spent sitting [50], whereas the other two studies assessed TV viewing [51, 53]. All three studies [50, 51, 53] reported no statistically significant associations between sedentary behavior and all-cause mortality. With regard to CRC-specific mortality, only one study, the CPS II Nutrition Cohort showed a statistically significant positive association between sedentary behavior and CRC-specific mortality (≥ 6 h vs. < 3 h/day sitting time: HR 1.62; 95% CI 1.07–2.44) [50].

Evidence from prospective observational studies has consistently suggested that higher physical activity after CRC diagnosis is associated with a lower risk of CRC-specific and all-cause mortality, but whether physical activity is causally related to CRC mortality remains unclear. A randomized controlled trial is currently ongoing to address whether aerobic physical activity after complement of adjuvant therapy improves survival. Based on a few studies, there is some evidence suggesting that excessive sedentary behavior after CRC diagnosis might be associated with increased CRC-specific mortality, but findings are less consistent than for leisure time physical activity.

For all-cause mortality, eight out of nine population-based studies [26, 28, 29, 31, 54‐57] reported increased all-cause mortality risk for smoking, of which six [26, 28, 29, 54, 55, 57] were statistically significant. Furthermore, the study in the adjuvant setting also reported a statistically significant increased all-cause mortality risk for smoking [61].

For CRC-specific mortality, five population-based studies [14, 29, 30, 57, 58] reported increased CRC-specific mortality risk for smoking, of which three [14, 29, 30] were statistically significant (Table 1). However, one study that reported results separately for men and women reported a statistically non-significant positive association among women for post-diagnosis smoking, while among men, a statistically non-significant inverse association was reported [56]. Furthermore, one study in the adjuvant setting also reported a statistically significant increased CRC-specific mortality risk for smoking [59].

For colon cancer recurrence, one study embedded in the trial N0147 [61] reported a statistically significant increased cancer recurrence risk for smoking, while CALGB 89803 [60] reported no association with smoking among stage III colon cancer patients treated with adjuvant chemotherapy.

Four population-based studies provided results on smoking cessation and mortality outcomes in CRC patients (Table 2). People who continued smoking after CRC diagnosis had a more than threefold increased risk of all-cause mortality (HR 3.46; 95% CI 1.69–7.10) compared to people who quit smoking after diagnosis [28]. Pre- to post-diagnosis smoking cessation was not statistically significantly associated with all-cause or CRC-specific mortality risk [29, 58, 79], although one of these studies reported lower mortality risk for those who quit smoking compared to those who continued to smoke [29].

Overall, evidence from observational studies has consistently suggested that smoking after CRC diagnosis increases the risk of CRC-specific and all-cause mortality. It seems plausible that smoking cessation would improve survival outcomes in CRC survivors, although direct evidence is limited.

Eleven population-based studies [16, 26, 32, 33, 48, 49, 62‐66], two studies from adjuvant chemotherapy trials [67, 68], and one study among metastatic patients [34] assessed the association of BMI at or after CRC diagnosis and CRC outcomes (Table 1). Furthermore, 21 additional studies in the adjuvant setting were included in a pooled analyses of patients enrolled in trials of adjuvant chemotherapy [69]. Moreover, an additional article with pooled analyses in the metastatic setting included data of 25 treatment trials [70].

For underweight (either BMI < 18.5 or 20 kg/m²), all population-based studies [16, 26, 32, 33, 48, 62‐65], the pooled analysis of studies in the adjuvant setting [69], and both publications in the metastatic setting [34, 70] reported higher all-cause mortality risk compared to normal weight individuals. The majority of these studies [26, 32, 34, 48, 62, 65, 69, 70] reported statistically significant results (Table 1). In the largest population-based study, ~ 3400 men and women diagnosed with stage I to III CRC from the Kaiser Permanente Northern California population, underweight at diagnosis was associated with a threefold increased all-cause mortality risk (HR 3.01; 95% CI 1.88–4.83) compared to normal weight [32]. However, most other studies report a 1.5- to 2-fold increased risk (Table 1). Generally, similar results were reported for CRC-specific mortality and cancer recurrence (Table 1).

For overweight (defined as BMI 25.0–24.9 kg/m²), all population-based studies [16, 26, 32, 33, 48, 49, 62, 64‐66] reported lower all-cause mortality risk compared to normal weight individuals, of which three were statistically significant [48, 49, 62]. However, studies in the adjuvant setting of a chemotherapy trial reported that overweight individuals had a similar all-cause mortality risk as normal weight individuals (Table 1). For metastatic patients participating in treatment trials, all-cause mortality risk was lowest at BMI 28 kg/m² [70], while overweight was associated with an increased all-cause mortality risk among a general population of patients diagnosed with metastatic disease (HR 1.23; 95% CI 1.03–1.46) [34]. Generally, similar results were reported for CRC-specific mortality and cancer recurrence (Table 1).

For obesity (BMI ≥ 30 kg/m²), none of the population-based studies [16, 26, 32, 33, 48, 49, 62, 64‐66] reported statistically significant associations with all-cause mortality. Nevertheless, the only study (Kaiser Permanente Northern California cohort) that reported on a separate group with class II or III obesity (BMI ≥ 35 kg/m²) reported a statistically significant increased all-cause mortality risk [32]. Within the adjuvant setting pooled analyses showed a modest increased all-cause mortality risk (HR 1.10; 95% CI 1.04–1.17) compared with normal weight [69]. Within the metastatic setting, both publications showed that obese individuals had a somewhat similar, or lower, all-cause mortality risk as normal weight individuals [34, 70]. Generally, similar results were reported for CRC-specific mortality and cancer recurrence (Table 1).

Four studies [48, 80‐82] reported on weight changes (Table 2). Two studies were population-based studies, a cohort from the Kaiser Permanente Northern California population [80] and an Australian cohort [48], and two studies were in the adjuvant setting, CALGB 89803 [81] and a cohort from the British Columbia Cancer Agency [82].

Large post-diagnosis weight loss (> 5 kg or ≥ 10%) was associated with a threefold increased all-cause and CRC mortality risk compared with stable weight in both population-based studies [48, 80]. Modest weight loss (2–4.9 kg or 5–9.9%) was also associated with increased all-cause and CRC mortality risk [48, 80], although only statistically significant in the Kaiser Permanente Northern California cohort [80]. In fact, the association between weight loss and mortality was present regardless of at-diagnosis BMI [80]. Large weight loss during adjuvant chemotherapy was associated with increased all-cause mortality and recurrence risk in a cohort from the British Columbia Cancer Agency [82], but not in CALGB 89803 [81].

Post-diagnosis weight gain was not associated with increased all-cause or CRC-specific mortality risk [48, 80, 81] or colon cancer recurrence [81, 82]. Furthermore, pre- to post-diagnosis weight loss or weight gain of >5 kg were both associated with a statistically significant 60% higher all-cause risk compared to stable weight [48].

Three population-based studies [35, 39, 71], two studies in the adjuvant setting [72, 73], and one study among metastatic patients [74] reported on post-diagnosis visceral adipose tissue and all-cause mortality (Table 1). Most of these studies were small (n = 62 to 339), except for the population-based cohort from the Kaiser Permanente Northern California population (n ~ 3200) [39••].

For all-cause mortality, all population-based studies [35, 39••, 71] reported statistically non-significant associations with visceral adipose tissue (Table 1). Both studies among patients treated with chemotherapy [72, 73] reported an increased all-cause mortality risk with high visceral adipose tissue, of which one was statistically significant [72]. The study among metastatic CRC patients [74] reported a statistically significant increased all-cause mortality risk for high visceral adipose tissue among patients treated with chemotherapy plus the angiogenesis inhibitor bevacizumab, but not among patients treated with chemotherapy only.

Four population-based studies [35, 38, 39••, 75], one study in the adjuvant setting [37] and three studies among patients with metastatic disease [36, 76, 77] reported on all-cause mortality (Table 1). Most of these studies were small (n = 67 to 339), except two population-based cohorts, from the Kaiser Permanente Northern California population (n ~ 3200) [39••] and from a single-institution hospital cohort that included stage I–IV patients [38].

Seven out of eight studies [36‐38, 39••, 75‐77] reported increased all-cause mortality risk for low skeletal muscle mass, of which five were statistically significant [37, 38, 39••, 75, 76] (Table 1). A meta-analysis, based on three small studies [75‐77], concluded that a low muscle mass was statistically significantly associated with a more than twofold increased all-cause mortality risk (HR 2.25; 95% CI 1.63–3.09) [20]. The only large population-based cohort with non-metastatic patients, from Kaiser Permanente Northern California, showed an almost 30% increased risk of overall mortality and 50% increased risk of CRC-specific mortality [39••].

One study among metastatic patients reported on loss of muscle mass during chemotherapy [36]. This study showed that ≥ 9% loss of muscle mass during chemotherapy was associated with a more than fourfold increased all-cause mortality risk (HR 4.47; 95% CI 2.21–9.05) [36].

Body fatness was studied most often by assessment of body mass index, while only few studies assessed other measures of body composition. Altogether, the results of studies across the three study categories (population-based, adjuvant, and metastatic setting) suggest a J- or L-shaped association between BMI and all-cause mortality or CRC-specific mortality risk. The risk of death was highest among patients who were underweight, while lowest risk was seen in patients with a BMI between 25 and < 30 kg/m². If obesity confers an additional mortality risk compared to normal weight or overweight patients remains uncertain. Nevertheless, the most recent meta-analysis of post-diagnosis BMI concluded that obesity was statistically significantly associated with a modest 8% increased all-cause mortality risk (HR 1.08; 95% CI 1.03–1.13) compared to normal weight, while no association was found between obesity and CRC-specific mortality [17••]. Weight loss in the first 2 years after diagnosis was consistently associated with increased mortality risk and this association was independent of BMI at CRC diagnosis. Currently, there are no intentional weight loss trials among CRC survivors that assessed mortality risk [86••] and no study that assessed the effect of weight loss after treatment was successfully completed. That being overweight, and in some studies even obese states, seem to be associated with improved survival compared to normal weight is called the “obesity paradox.” The obesity paradox could be explained by several methodological issues, including the crudeness of BMI as a measure of body fatness, especially in a cancer patient population where loss of weight and lean body mass is a strong adverse factor [87].

Other measures used to study the association between body composition and CRC outcomes were visceral adipose tissue and muscle mass quantified from CT images; studies with other measures, such as waist circumference, are currently lacking. There is only limited evidence that visceral adiposity increased mortality risk. Across study categories, studies had mixed results. Only in the adjuvant setting, two small studies consistently showed increased all-cause mortality risk with higher visceral adipose tissue. Even though quantification of adipose tissue from CT scans is regarded as a more precise measure of adiposity than BMI, the usefulness of single-slice analysis might be limited [88]. On the other hand, evidence consistently shows that low muscle mass is associated with reduced survival, although each study used other cut points to define low muscle mass. The notion that the association between overweight and lower mortality is due solely to methodologic biases is refuted by results from the only large population-based study among non-metastatic CRC patients with available data for both BMI and body composition [39••]. Within the overweight BMI range between 25 and < 30 kg/m², body composition appeared to explain why a BMI higher than normal is associated with the lowest mortality. The majority (78%) of patients in the overweight group had adequate muscle mass, while less than half (43%) of the patients with a normal BMI had adequate muscle mass. Furthermore, the obesity paradox could also be explained by clinical issues [87], such as metabolic health. One study at Kaiser Permanente investigated the combination of obesity and metabolic health and concluded that mortality risk was statistically significantly increased in obese patients with the metabolic syndrome, but not in metabolically healthy obese patients, compared with metabolically healthy non-obese patients [89].