Dietary Intake in Association with All-Cause Mortality and Colorectal Cancer Mortality among Colorectal Cancer Survivors: A Systematic Review and Meta-Analysis of Prospective Studies

Simple Summary Given that an extensive range of dietary factors has not been investigated among colorectal cancer (CRC) survivors to date, we carried out a systematic review and meta-analysis to determine the effects of both prediagnostic and postdiagnostic dietary intake on all-cause mortality and CRC-specific mortality among CRC survivors. In total, 45 studies were included in the final analysis of 35 food items, 8 macronutrients, 27 micronutrients, 2 dietary patterns, and 13 dietary indexes in association with all-cause mortality and CRC-specific mortality. We found that an unhealthy dietary pattern increased the risks of both all-cause mortality and CRC-specific mortality. The role of prediagnostic and postdiagnostic intake such as macronutrients and fatty acids could be different in the risk of all-cause mortality. Overall, comprehensive evidence for the effect of substantial numbers of prediagnostic and postdiagnostic dietary items on mortality outcomes is reported in this study. Abstract We carried out a systematic review and meta-analysis to determine the effects of both prediagnostic and postdiagnostic dietary intake on all-cause mortality and CRC-specific mortality among CRC survivors. An extensive search of PubMed and Embase was conducted to identify eligible studies. We applied a random-effects model to estimate the pooled relative risks (RRs)/hazard ratios (HRs) and their 95% confidence intervals (CIs). As a result, a total of 45 studies were included in the final analysis. Pooled effect sizes from at least three study populations showed that whole grains and calcium were inversely associated with all-cause mortality, with RRs/HRs (95% CIs) of 0.83 (0.69–0.99) and 0.84 (0.73–0.97), respectively. In contrast, a positive association between an unhealthy dietary pattern and both all-cause mortality (RR/HR = 1.47, 95% CI = 1.05–2.05) and CRC-specific mortality (RR/HR = 1.52, 95% CI = 1.13–2.06) was observed among CRC survivors. In the subgroup analysis by CRC diagnosis, prediagnostic and postdiagnostic dietary intake such as carbohydrates, proteins, lipids, and fiber were observed to have different effects on all-cause mortality. Overall, an unhealthy dietary pattern increased the risks of both all-cause mortality and CRC-specific mortality. The role of prediagnostic and postdiagnostic intake of dietary elements such as macronutrients and fatty acids could be different in the risk of all-cause mortality.

. Flow chart of the study selection. The flow chart shows the process used to select prospective studies for the systematic review and meta-analysis of the association between dietary intake and allcause mortality and colorectal cancer (CRC) mortality among CRC survivors. RR, relative risk; HR, hazard ratio.

Main Analysis
Possible publication bias when four or more studies were available was assessed by Begg's funnel plot and Egger's test ( Figure 2). Publication bias was observed for the association of dietary whole grain (p = 0.02) or alcohol (p = 0.04) and CRC death. Possible publication bias when four or more studies were available was assessed by Begg's funnel plot and Egger's test ( Figure 2). Publication bias was observed for the association of dietary whole grain (p = 0.02) or alcohol (p = 0.04) and CRC death.

Subgroup Analysis
The subgroup analysis of the associations between prediagnostic dietary intake (27 food items, 8 macronutrients, 27 micronutrients, 2 dietary patterns, and 13 dietary indexes) and mortality among CRC survivors is presented in Table 2. The observed pattern of a negative association between fruits and all-cause mortality and of a positive association between grilled foods and CRC-specific mortality was similar to that in the results of the main analysis.

Subgroup Analysis
The subgroup analysis of the associations between prediagnostic dietary intake (27 food items, 8 macronutrients, 27 micronutrients, 2 dietary patterns, and 13 dietary indexes) and mortality among CRC survivors is presented in Table 2. The observed pattern of a negative association between fruits and all-cause mortality and of a positive association between grilled foods and CRC-specific mortality was similar to that in the results of the main analysis. The subgroup analysis for the associations between postdiagnostic dietary intake (15 food items, 8 macronutrients, 2 micronutrients, 2 dietary patterns, and 9 dietary indexes) and mortality among CRC survivors is presented in Table 3. The observed pattern in the association between refined grains, whole grains, dark fish, coffee consumption, ACS score, or glycemic load and all-cause mortality was similar to that of the results from the main analysis.  A comparison of the roles of prediagnostic and postdiagnostic intake in terms of all-cause mortality is presented in Figure 3. Unhealthy prediagnostic and postdiagnostic patterns (RR/HR = 1.33, 95% CI = 1.09-1.62 and RR/HR = 1.47, 95% CI = 1.05-2.05, respectively), the insulin index (RR/HR = 1.32, 95% CI = 1.02-1.71 and RR/HR = 1.89, 95% CI = 1.22-2.91, respectively), and insulin load (RR/HR = 1.33, 95% CI = 1.03-1.72 and RR/HR = 2.30, 95% CI = 1.36-3.87, respectively) were consistently associated with an increased risk of all-cause mortality. Additionally, the prediagnostic and postdiagnostic ACS scores were consistently associated with a decreased risk of all-cause mortality, with RRs/HRs (95% CIs) of 0.78 (0.69-0.94) and 0.62 (0.43-0.89), respectively. The all-cause mortality was 73% lower in the participants with a high consumption of prediagnostic proteins (RR/HR = 0.27, 95% CI = 0.12-0.63) but 24% higher in those with high consumption of postdiagnostic proteins (RR/HR = 1.24, 95% CI = 1.03-1.49). The prediagnostic, but not postdiagnostic, intake of carbohydrates, proteins, lipids, saturated fatty acids (SFAs), and monounsaturated fatty acids (MUFAs) and the (modified and/or alternative) Mediterranean Diet score (MED/aMED) were associated with decreased risks of all-cause mortality, with RRs/HRs (95% CIs) of 0.32 (0.14-0. 76  A comparison of the roles of prediagnostic and postdiagnostic intake in terms of CRC-specific mortality is presented in Figure 4. Consistently, significant associations were observed between the prediagnostic and postdiagnostic intake and coffee intake (RR/HR=0. 46

Discussion
The current systematic review and meta-analysis of 45 prospective studies investigated the association between dietary intake and the risk of all-cause mortality and CRC-specific mortality among CRC survivors. Our findings indicated diverse patterns of the associations between all-cause mortality or CRC-specific mortality and each food item, nutrient, dietary pattern, or index. Additionally, similar associations with mortality before and after the CRC diagnosis were observed with red and processed meat, sugar-containing and sweet products, total dairy, coffee, alcohol, calcium, vitamin D, an unhealthy pattern, Dietary Approaches to Stop Hypertension (DASH) score, and insulin index or load among CRC survivors; however, different associations were found with the intake of whole grain, milk, the macronutrients of carbohydrates, proteins, lipids, SFA, MUFA, PUFA, and fiber intake; a prudent pattern; and dietary indexes, such as the ACS, HEI/aHEI, MED/aMED, and DII/eDII.
In this study, the risk of all-cause mortality varied by food item, nutrient, dietary pattern, and index. A significantly reduced risk of all-cause mortality was associated with the highest intake of fruits, whole grains, dark fish, coffee, and calcium and the lowest ACS score, while an unhealthy pattern and the highest insulin index were associated with an increased risk of all-cause mortality. Regarding CRC-specific mortality, coffee intake and an unhealthy dietary pattern were associated with a lower and higher risk of CRC-specific mortality, respectively. Our results are consistent with the findings of recent meta-analyses of prospective studies; inverse associations were observed between all-cause mortality and the intake of whole grains [61,62], fruit [62,63], fish [62], coffee [64], and calcium [65], and mostly nonlinear relationships were observed. However, in contrast to our results, in the previous meta-analyses, there were negative associations between all-cause mortality and vegetable [62,63] or nut [62] consumption; the HEI, aHEI, and DASH scores [66]; and a prudent/healthy dietary pattern [67]. Additionally, there was a positive association between all-cause mortality and red and processed meat intake [62] but no association with soy consumption [68] or a Western/unhealthy dietary pattern [67]. Of these dietary components, patterns, and indexes, a high intake of total calcium, dietary fiber, whole grains, and soy and the combination of the three indexes (HEI, aHEI, and DASH) were associated with a reduced risk of CRC [66,[68][69][70][71]. Furthermore, a Western dietary pattern was associated with an increased risk of CRC-specific mortality [7]. Specifically, two previous meta-analyses of dietary factors and CRC-specific mortality were targeted to cancer survivors [7,66]; a Western dietary pattern had an adverse effect on CRC-specific mortality (RR = 1.55, 95% CI = 1.13-2.13, I 2 = 35%), but no association was found with dairy or meat consumption or a prudent/healthy dietary pattern among 209,597 cancer survivors [7]. However, all three indexes of dietary quality combined (HEI, aHEI, and DASH) had a protective effect on CRC-specific mortality in a previous meta-analysis of seven studies involving cancer survivors (RR = 0.77, 95% CI = 0.73-0.81, I 2 = 0%) [7]. Nevertheless, no specific information on guidelines regarding various dietary factors are available to date for overall cancer survivors, including CRC survivors. Concerning different intervention strategies that would be needed according to the cancer type, dietary recommendations for patients with CRC should be considered at the time of diagnosis.
Changes in health-related behaviors after diagnosis and treatment, such as dietary patterns or physical activity, have been investigated among CRC survivors. Van Zutphen et al. [4] reported that CRC survivors consumed significantly fewer sugary drinks (−45 g/day) and less red and processed meat (−62 g/week) at 2 years after diagnosis. In this study, similar patterns of the risk of CRC-specific mortality, rather than all-cause mortality, were significantly shown with prediagnostic and postdiagnostic dietary intake among CRC survivors, including coffee consumption, insulin index, and insulin load. The highest intake of coffee was associated with a lower risk of CRC-specific mortality both before and after the CRC diagnosis. However, the highest insulin index or load group was found to have a higher risk of CRC-specific mortality. Consistent with our results, the Nurses' Health Study and Health Professional Follow-up Study found that 1599 CRC patients with a stage of I to III who maintained coffee intake more than two cups a day after the CRC diagnosis had a lower risk of CRC-specific mortality (HR = 0.63, 95% CI = 0.44-0.89) than those consuming coffee intake below two cups a day before and after diagnosis [27]. A slightly weaker inverse association between coffee intake and all-cause mortality has been observed (HR = 0.71, 95% CI = 0.60-0.85; maintaining ≥ 2 cups/day vs. maintaining < 2 cups/day) [27]. Interestingly, in a subgroup analysis of the insulin load or index, which were defined as risk factors for both all-cause mortality and CRC-specific mortality in this study, a lower risk of CRC-specific mortality with coffee intake was observed only in the lower category of insulin load (HR = 0.83, 95% CI = 0.70-0.98) despite the null association for the dietary insulin index [27]. The insulin index can be less reflective of the long-term effect on CRC-specific mortality than the insulin load [72], and its role in CRC-specific mortality among survivors may result from the combination of dietary items they consumed, such as carbohydrates, proteins, and fiber, or the strongest item among these dietary components.
In this study, the all-cause mortality and CRC-specific mortality were consistently and positively affected by an unhealthy dietary pattern depending on the CRC prediagnostic and postdiagnosis status. Regarding the prediagnostic unhealthy dietary patterns, which were defined as a high-sugar or processed meat patterns, there were significant positive associations with all-cause mortality (RR = 1.33, 95% CI = 1.09-1.62) and CRC-specific mortality (RR = 1.34, 95% CI = 1.01-1.78). Furthermore, regarding the postdiagnostic unhealthy patterns, namely, the Western dietary pattern reported in three studies, significant positive associations were observed with both all-cause mortality (RR = 1.47, 95% CI = 1.05-2.05) and CRC-specific mortality (RR = 1.69, 95% CI = 1.09-2.64). These results could be attributable to the combination of dietary factors comprising the unhealthy dietary pattern, rather than each dietary component; however, little knowledge regarding its definition among CRC survivors is limited. On the basis of the three articles included in the final analysis, an unhealthy pattern was commonly characterized by high intake of red and processed meats and refined grains, as well as additionally included sweets, desserts, and high-fat dairy products [4]; eggs, solid fats, and salty snacks [4]; or desserts, high-fat dairy products, and French fries [4]. Further studies are needed to investigate the effect of a Western dietary pattern on mortality on the basis of a clarified combination of dietary components.
A large prospective study of 31,456 deaths during 9 years of follow-up observed the protective effect of dietary fiber intake on reducing all-cause mortality among both men (RR = 0.78, 95% CI = 0.73-0.82) and women (RR = 0.78, 95% CI = 0.73-0.85) [73], which was consistent with our findings of postdiagnostic fiber intake. However, fiber intake was associated with a decreased risk of cancer death among men only (RR = 0.83, 95% CI = 0.76-0.92) and not among women (RR = 0.96, 95% CI = 0.85-1.08) [73]. In contrast, prediagnostic fiber intake was consistently not associated with all-cause mortality or CRC-specific mortality among CRC survivors [20,50,56]. Although the amount of prediagnostic and postdiagnostic fiber consumption did not differ among Norwegian women with a CRC diagnosis [50,74], a per 5 g/day increase in fiber intake after the CRC diagnosis was associated with 14% and 18% decreased risks of all-cause mortality and CRC-specific mortality, respectively [44]. However, the role of carbohydrates in the risk of CRC has been controversial [75][76][77][78]. A pooled analysis of 17 observational studies found that a high consumption of carbohydrates was not associated with CRC and that it did not differ by colon or rectal cancer [75]. Despite the small number of individual studies in our analysis, a significant positive association was observed in the effect of carbohydrate intake on CRC-specific mortality, but not all-cause mortality, leading to the result of postdiagnostic carbohydrate intake. Considering that a high intake of carbohydrates could result in a higher dietary insulin index and load [72], further studies are warranted to specify the effects on CRC-specific mortality by the source of carbohydrate intake following the CRC diagnosis. There has also been limited evidence regarding the role of protein intake in CRC-specific mortality, and only one null association has been reported thus far [79]. Therefore, changes in many dietary factors before and after a disease diagnosis [80] may represent an important factor contributing to mortality outcomes. A recent systematic review of diet and physical activity related to the quality of life of CRC survivors highlighted the importance of a healthy diet along with physical activity for reducing the risk of recurrence [81], suggesting adherence to a healthy lifestyle is necessary for patients with CRC to improve their quality of life and prolong their lives. Regarding a prudent pattern, null associations between all-cause mortality and both the CRC prediagnostic and postdiagnosis intake were observed. A protective effect on CRC-specific mortality was observed only with a postdiagnostic prudent pattern consumption.
To the best of our knowledge, this is the first study to analyze all the available data of both prediagnostic and postdiagnostic dietary intake in association with all-cause mortality and CRC-specific mortality among CRC survivors. The current systematic review and meta-analysis included only prospective studies, which are assumed to have a higher level of evidence and are less susceptible to recall bias and selection bias than retrospective studies [82]. Additionally, the methodological quality assessment showed that all the studies included in the final analysis were of high quality. Furthermore, in all studies concerning dietary intake and mortality outcomes, adjustments were made for major risk factors, suggesting that the findings were more reliable. Despite its strengths, this study has some limitations. First, the association between several dietary items and the mortality outcomes was reported from a single study only, which did not allow us to calculate the pooled estimate of various study populations. Second, the cut-off level of the highest and lowest quantiles was generally unclear and nonuniform across the individual studies, suggesting that both overestimations and underestimations could have occurred [83]. Third, the intrinsic limitation of potential biases from individual studies included in the final analysis might affect our pooled estimates. Whether the dietary habits observed at baseline were maintained during follow-up was unclear. The subjects who were aware of the important role of a healthy diet might change their consumption behavior after the cancer diagnosis [84]. Finally, the dietary intake in the included studies was obtained by using an FFQ, which is completed by the participants themselves or via an interview [82]. Both web-based and print-based FFQ have been reported to share measurement error [85], which may affect the power of detecting the diet-mortality association.

Conclusions
In summary, this study provided comprehensive evidence of the effect of all prediagnostic and postdiagnostic dietary items on mortality outcomes. Overall, unhealthy consumption patterns, including a Western diet, the intake of processed meat, or high-sugar dietary patterns, was found to increase the risks of both all-cause mortality and CRC-specific mortality. Prediagnostic and postdiagnostic intake of whole grain, carbohydrates, proteins, lipids, SFAs, MUFAs, PUFAs, and fiber played different roles in the risk of all-cause mortality.
Supplementary Materials: The following are available online at http://www.mdpi.com/2072-6694/12/11/3391/s1: Table S1: Search strategy. Table S2: Inclusion and exclusion criteria. Table S3: Baseline characteristics of the studies included in the final systematic review and meta-analysis. Table S4: Newcastle-Ottawa quality assessment scale of the cohort studies included in the systematic review and meta-analysis. Table S5: Assessment of attrition bias in the individual studies.