Association of dietary carbohydrate intake, glycemic index, and glycemic load with colorectal cancer risk in adults: a systematic review and dose-response meta-analysis of prospective cohort studies

Abstract

Background and objectives

Carbohydrate consumption is a key factor in controlling insulin secretion; however, the effects vary by the glycemic index (GI) and glycemic load (GL), which might be associated with the risk of colorectal cancer. Due to conflicting results regarding their impact on colorectal cancer and considering that previous meta-analyses lacked dose-response analyses for colon and rectal cancer, this comprehensive dose-response meta-analysis was undertaken to explore the association between dietary carbohydrates, GI, and GL with the risks of colorectal, colon, and rectal cancers.

Method

A comprehensive search of online databases (PubMed, Scopus, Web of Sciences) up to November 2024 identified 23 studies with 2,019,665 participants. Random-effect models were used to calculate pooled effect sizes, and dose-response analyses assessed linear and nonlinear associations. We used the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) form.

Result

The results of this study revealed a marginal increase in colorectal cancer risk associated with a higher GI (RR: 1.08, 95% CI: 1.00 − 1.16, P = 0.04). However, in subgroup analyses, higher GI was significantly associated with an increased risk of colorectal cancer among men, in cohorts with follow-up durations shorter than 10 years, and in studies conducted in US populations. On the other hand, subgroup analyses indicated an inverse significant association between carbohydrate intake and colorectal cancer in specific subgroups, including men, women, studies conducted in non-US populations, and cohorts with follow-up durations shorter than 10 years. Furthermore, neither linear nor nonlinear dose-response analyses indicated a significant association between exposures and outcomes.

Conclusions

These results suggest that a higher GI may be associated with an increased risk of colorectal cancer. Nevertheless, additional studies are required to establish a more definitive connection between carbohydrate intake and colorectal cancer risk.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12885-025-15466-1.

Introduction

Colorectal cancer (CRC) was among the top three most diagnosed cancers globally in 2020, and it stood as the second leading cause of death attributed to cancer [1]. Despite considerable improvements in both screening techniques and treatment options, CRC remains a growing global burden, underscoring the need to identify and modify risk factors associated with its development [2]. CRC is characterized as a multifactorial condition influenced by both genetic predisposition and environmental factors. Among the modifiable environmental determinants, dietary intake plays a significant role [3–5]. Carbohydrates are a key energy source, making up a large part of the diet; the type and quantity affect blood glucose levels and insulin secretion [6, 7]. High carbohydrate intake raises postprandial blood glucose, boosting insulin secretion and IGF-1 (Insulin-like Growth Factor-1) levels. This powerful mitogen prevents apoptosis and stimulates cell growth [8–10]. Certain foods rapidly raise blood glucose, releasing more insulin. Jenkins introduced the Glycemic Index (GI) in 1981 as a system that ranks foods based on their blood glucose–raising potential relative to white bread or glucose [11]. The GI measures carbohydrate quality but not quantity. However, GL (Glycemic Load) divides the GI by 100 and includes both carbohydrate quality and quantity (in grams) [12].

Diets containing foods with a high GI and GL cause postprandial blood sugar spikes, increasing insulin secretion by pancreatic β-cells, thereby facilitating glucose regulation in the bloodstream [13]. Over time, this recurrent response can lead to the development of insulin resistance [14]. Low-grade inflammation commonly coexists with insulin resistance and can further increase systemic inflammatory burden as insulin resistance worsens [15]. This chronic inflammatory state ultimately contributes to cellular damage and tumorigenesis [16]. Higher levels of inflammation-related indicators like CRP) C-Reactive Protein (are correlated with a greater likelihood of developing colorectal cancer, highlighting the interplay between inflammation and cell cycle dysfunction [17]. The relationship between carbohydrate consumption, GI, GL, and CRC has been examined in several meta-analyses, yet the outcomes have not been conclusive. Aune et al. were the first to publish a meta-analysis in 2012 focusing on the relationship between carbohydrate intake, GI, GL, and CRC. In their study, these dietary parameters did not significantly affect CRC risk [18]. In 2015, GI and GL were examined with numerous cancer forms, including CRC. GL and CRC were not significantly associated when comparing the highest and lowest consumption groups, while GI was [19]. Huang et al., in a 2017 meta-analysis, concluded that total carbohydrate intake was not associated with CRC risk. However, subgroup analysis showed that men with higher carbohydrate consumption may have a higher CRC risk [20]. In 2019, the latest meta-analysis found no connection between carbohydrate intake or GL and CRC. However, it showed that a higher dietary GI may increase CRC risk [21].

This systematic review and meta-analysis were conducted as a comprehensive approach to resolving major gaps in the understanding of the topic. Firstly, the need for an updated synthesis became evident following the recent publication of three large-scale cohort studies [22–24], which were not included in previous analyses. Secondly, several eligible studies were not included in earlier meta-analyses, which limited the comprehensiveness of their conclusions. Thirdly, prior investigations had not systematically assessed the potential differential effects between colon and rectal cancers through independent linear and nonlinear dose-response analyses. Given these considerations, a comprehensive systematic review and meta-analysis are essential to clarify the relationships between dietary carbohydrate intake, GI, and GL and their associations with CRC, with specific attention to potential site-specific and dose-response effects.

Method

The protocol for this systematic review and meta-analysis was registered with PROSPERO (CRD42024597572), and no amendments were made to the original protocol. The study was designed and conducted in full accordance with the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines [25].

Search strategy

We conducted a comprehensive literature search across the online databases Scopus, PubMed, and Web of Science (WOS) up to November 2024. The complete search terms are documented in Supplementary Table 1. No restrictions were applied regarding the publication date or language. All search results were imported into EndNote software, where duplicate entries were removed. Two investigators (MMM and NAH) independently screened the titles and abstracts, and any discrepancies were resolved through consultation with LA. Additionally, the reference lists of relevant articles were examined to ensure no studies were overlooked. An additional manual search was performed in Google Scholar using the primary keywords to ensure the inclusion of relevant studies that might have been missed during the database searches.

Inclusion criteria

Eligibility criteria for article selection included the following: (1) observational studies utilizing a prospective cohort design; (2) investigations examining the association between dietary carbohydrate intake, GI, and GL as exposure variables, and the risk of colorectal, colon, or rectal cancer as the outcome; (3) publications conducted among adults aged 18 years or older; and (4) studies providing relative risk measures—such as odds ratios (OR), risk ratios (RR), or hazard ratios (HR), accompanied by 95% confidence intervals (CI) to explore these associations. If multiple articles were based on the same dataset, priority was given to the most recent publication; otherwise, the study with the largest sample size or the longest follow-up period was selected.

Exclusion criteria

The exclusion criteria encompassed meta-analyses, systematic reviews, narrative reviews, umbrella reviews, and other types of review articles. Additionally, letters, comments, short communications, ecological studies, and animal research were omitted. Studies employing cross-sectional, nested case-control, case-cohort, or case-control designs were also excluded. Furthermore, studies lacking critical data required for meta-analysis, such as relative risk (RR), hazard ratio (HR), or data necessary for their calculation, were not considered.

Data extraction

Two researchers (MMM and NAH) independently extracted and verified data from the included articles. The following characteristics of the identified studies were recorded: first author’s name, year of publication, study location, sex, age (mean or range), sample size, number of cases, cancer outcomes (colorectal cancer, colon cancer, or rectal cancer), type of exposure (GI, GL, or carbohydrates), methods used for dietary assessment and cancer diagnosis, duration of follow-up, risk estimates (RR, hazard ratio, or OR with 95% CI), and confounding variables adjusted for in the statistical analysis. If a study reported findings separately by sex, it was treated as two distinct studies. When multiple risk estimates were available, the fully adjusted effect sizes were extracted.

Quality assessment

Quality assessment of the included studies in this meta-analysis was performed using the Newcastle-Ottawa Scale (NOS), which is particularly suited for prospective cohort studies [26]. Each study was assigned a maximum of 9 points: 4 points for participant selection, 2 points for comparability, and 3 points for the assessment of outcomes and follow-up duration. The individual study scores, along with detailed scores for each section, are available in Supplementary Table 2.

Statistical analysis

Relative risks (RRs), and hazard ratios (HRs), along with their 95% confidence intervals (CIs), for comparison of the highest versus lowest categories of dietary carbohydrate intake, GI, and GL were used to calculate log RRs, and HRs with standard errors. A random effect model was used for the analysis, with heterogeneity assessed using the Q statistic and Heterogeneity was considered significant if the P-value Q-test was < 0.05 or if I² exceeded 50%. Pre-specified subgroup analyses were performed to examine potential sources of heterogeneity. In subgroup analysis, studies were categorized based on sex (male participants/female participants/both), study location (United States/non-United States), and duration of follow-up (≥ 10 years vs. <10 years). Heterogeneity between subgroups was assessed using a fixed-effects model. To assess publication bias, funnel plots were visually examined, and Egger’s test was conducted for formal statistical evaluation. Sensitivity analysis was also undertaken using a fixed-effects model, involving the sequential exclusion of individual prospective cohort studies to determine their impact on the pooled risk estimate.

Utilizing the dose-response meta-analysis method proposed by Greenland and Orsini, this study investigated the association between dietary carbohydrate, GI, and GL and the risk of colorectal, colon, and rectal cancer [27, 28]. Both linear and nonlinear models were employed in this analysis. A linear dose-response analysis was performed using a generalized least squares regression method to estimate the study-specific slope. The eligible study reported the number of participants and cases, person-years of follow-up, median/mean of exposure levels (dietary carbohydrate intake, GI, and GL), adjusted RRs, and 95% CI for each category. For studies that reported carbohydrate intake as a percentage of energy, the values were converted to grams per day (g/day). For glycemic index (GI) and glycemic load (GL), we used the reported values and, where necessary, calculated the average of the upper and lower bounds for each category. In cases where the highest category was open-ended, the interval length was assumed to be the same as the adjacent category. Restricted cubic splines with three knots at the 10th, 50th, and 90th percentiles of the exposure distribution were used in the non-linear dose-response analysis to flexibly model potential non-linearity. To evaluate the significance of non-linearity, the coefficient of the second spline was examined to determine if it equaled zero. STATA version 17 was utilized for all analyses, with a p-value below 0.05 considered indicative of statistical significance.

Result

As illustrated in Fig. 1, the initial search identified 4782 studies. Among these, 663 were duplicate studies, and 4,085 were deemed irrelevant as they did not meet the inclusion criteria; therefore, they were excluded. Upon full-text review of the remaining 34 studies, 11 were excluded for the following reasons: 3 studies were excluded due to their case-control design [29–31], One study reported the association between carbohydrate quality and colorectal cancer, rather than quantity [32], 6 studies focused on the relationship between food groups, dietary pattern or diet without reporting carbohydrate intake [33–38] and one study investigated the association between diet and colorectal cancer mortality [39]. Finally, we included 23 cohort studies in the meta-analysis, 17 of which examined the association with carbohydrate intake, involving 1,950,608 participants and 19,677 cases of colorectal cancer [22, 24, 40–54], 13 studies including 1,380,215 participants and 15,305 colorectal cancer cases for GI [23, 45–49, 52–58] and 15 studies with 1,683,046 participants and 18,007 cases of colorectal cancer examined the association with GL [23, 44–50, 52–58].

Fig. 1.

Flow diagram of study selection

Characteristics of the included studies

Table 1 presents the characteristics of the cohort studies considered. The studies enrolled between 3,184 and 368,792 participants, aged 18 to 79 years. Overall, 2,019,665 participants were enrolled across the 23 publications included. During the follow-up periods, which ranged from 3 to 21.9 years, the total number of colorectal cancer cases was 32,333, with 16,205 cases of colon cancer and 5,932 cases of rectal cancer, all identified through national or regional cancer registries using standardized diagnostic procedures. 10 studies included only women, one study included only men, and six of the remaining studies reported relative risks for women and men separately (Table 1). 12 studies were performed in the United States, 3 in Asian countries, 6 in European countries, and 2 in Canada. Dietary carbohydrate intake, GI, and GL were assessed using FFQ in 20 studies, food recall in 2, and food records in one study. The NOS scores of the studies ranged from 6 to 9, with 10 studies scoring 8 or higher, indicating high quality. The scoring details for each section are presented separately in Supplementary Table 2. Most studies adjusted for several factors, including age (n = 20), energy intake (n = 22), BMI (n = 18), physical activity (n = 14), and alcohol consumption (n = 13), as well as other dietary variables such as processed meat, folate, calcium, and magnesium (n = 12). According to the NOS score, the quality scores of the included studies ranged from 6 to 9. Of these, 11 studies were rated ≥ 8, reflecting high methodological rigor, and were therefore classified as high-quality studies [23, 44, 46, 51, 53–58].

Table 1.

Characteristics of prospective cohort studies included in the current meta-analysis on the association between dietary carbohydrate intake, glycemic index, and glycemic load with colorectal cancer risk in adults

Author and publication year	Country	Age (years)	sex	Follow-up (years)	Exposure	Exposure assessment	Outcome	Sample size (F/M)	case	Quality score	adjustments
Bostick et al.,1994 [40]	USA	55–69	F	5	CHO	FFQ-127item	Colon cancer	35,215	212	7	Age-energy -height-parity-vitamin E intake- vitamin A intake
Chyou et al.,1996 [41]	USA	65–74	M	21.9	CHO	24 dietary recalls	Colon cancer	7954	330	6	Age
							Rectal cancer		123
Kato et al.,1997 [42]	USA	34–65	F	7.1	CHO	FFQ-70 item	Colorectal cancer	14,727	100	7	Energy-age-place-education
Nagata et al.,2001 [43]	JAPAN	≥ 35	F/M	3	CHO	FFQ-169 item	Colorectal cancer	F:15,754 M:12,607	F:98 M:181	7	Age- energy- year of smoking-alcohol
Terry et al.,2003 [44]	CANADA	40–59	F	16.5	CHO	FFQ	Colorectal cancer	49,124	616	8	Age-energy-BMI-study center-PA-oral contraceptive-hormone replacement therapy-parity-alcohol-red meat- folic acid-red meat- folic acid
							Colon cancer		436
							Rectal cancer		180
Oh et al.,2004 [59]	USA	30–55	F	18	CHO GI GL	FFQ-116 item	Colorectal cancer	34,428	1715	7	Age-BMI-smoking-alcohol-family history-year of endoscopy-aspirin use-menopausal status-PA-energy-total fiber-postmenopausal hormone use
							Rectal cancer		504
Higginbotham et al.,2004 [45]	CANADA	≥ 45	F	7.9	CHO GI GL	FFQ-131 item	Colorectal cancer	38,451	174	7	Age-BMI-oral contraceptive use-postmenopausal hormone use-family of history of colorectal cancer-smoking-alcohol-PA-energy-nsaid-folate-calcium
Michaud et al.,2005 [46]	USA	40–75	F/M	14	CHO GI GL	FFQ-61 item	Colorectal cancer	51,529	M:683 F:1096	8	Age-BMI-family of history of colorectal cancer-smoking-alcohol-PA-energy-nsaid-folate-calcium-height-intake of processed meat and cereal fiber
							Colon cancer		M:552 F:858
							Rectal cancer		M:131 F:238
McCarl et al.,2006 [55]	USA	55–69	F	15	GI GL	FFQ-127 item	Colorectal cancer	35,197	954	8	Age-energy-PA-multivitamin use-diabetes-smoking-BMI-waist/hip ratio
Larsson et al.,2006 [47]	SWEDEN	40–76	F	15.7	CHO GI GL	FFQ-67 item	Colorectal cancer	61,433	870	7	Education-BMI-energy-alcohol-folate-calcium-cereal fiber-red meat-magnesium
							Colon cancer		594
							Rectal cancer		283
Weijenberg et al.,2007 [56]	NETHERLAND	55–69	F/M	11.3	GI GL	FFQ-150 item	Colorectal cancer	120,852	M:1082 F:755	8	Energy-age-BMI-family history of colon cancer-alcohol-education-PA-calcium-energy-meat
							Colon cancer		M:674 F551
							Rectal cancer		M:280 F:138
Strayer et al.,2007 [48]	USA	MEAN: 61.9	F	8.5	CHO GI GL	FFQ-62 item	Colorectal cancer	45,651	490	6	Age-energy-NSAIDs-smoking-menopausal hormone use-calcium-fiber-BMI-screen for Colorectal cancer
Kabat et al.,2008 [49]	USA	50–79	F	7.8	CHO GI GL	WHI-FFQ	Colorectal cancer	158,000	1476	7	Age-education-smoking-BMI-hormone replacement therapy-family history-PA-energy- fiber-calcium
							Colon cancer		1149
							Rectal cancer		303
George et al.,2008 [57]	USA	50–71	F/M	8	GI GL	FFQ-124 item	Colorectal cancer	F:15,215 M:33,203	F:1457 M:3031	8	Age-race-education-marital status-PA-smoking-energy-alcohol-BMI-family history
Howarth et al.,2008 [50]	USA	45–75	F/M	8	CHO GL	QFFQ	Colorectal cancer	F:105,106 M:85,898	F:3864 M:4791	7	Age-family history-race-smoking-BMI-energy-multivitamin use-NSAIDs-folate-red meat-vitamin d – calcium- fiber
							Colon cancer		F:3008 M:3430
							Rectal cancer		F:818 M:1208
Ruder et al.,2011 [51]	USA	50–71	F/M	10	CHO	FFQ-124 item	Colon cancer	297,797	2794	8	Energy-sex-BMI-race-education-PA-smoking-NSAIDs-hormone replacement therapy-history of colon cancer
							Rectal cancer		979
Li et al.,2011 [52]	CHINA	40–70	F	9.1	CHO GI GL	FFQ	Colorectal cancer	73,061	475	7	Energy-age-education-BMI-PA-family history of colorectal cancer-hormone replacement therapy
							Colon cancer		287
							Rectal cancer		188
Sieri et al.,2014 [53]	ITALY	≥ 18	F/M	11.7	CHO GI GL	FFQ-150 item	Colorectal cancer	47,749	421	8	Energy-sex-PA-education-smoking-BMI-alcohol-calcium-folate-fiber
							Colon cancer		314
							Rectal cancer		107
Abe et al.,2016 [58]	JAPAN	40–69	F/M	12.5	GI GL	FFQ	Colorectal cancer	F:38,941 M:34,560	M:889 F:579	9	Age-area-alcohol-smoking-BMI-PA-menopausal status-exogenous female hormones-energy-calcium-vitamin B6-folate-vitamin B12- magnesium-vitamin D3
							Colon cancer		M:595 F:421
							Rectal cancer		M:294 F:158
Makarem et al.,2017 [54]	USA	Mean: 54.4	F/M	13.1	CHO GI GL	FFQ-126 item	Colorectal cancer	3184	68	8	Age-energy-sex-smoking-alcohol-red meat-fiber-processed meat
Papadimitriou et al.,2021 [24]	EUROPE	35–70	F/M	8	CHO	FFQ-150 item	Colorectal cancer	386,792	5069	7	Energy-smoking-BMI-PA-education-diabetes history-age-sex
Debras et al.,2022 [23]	FRANCE	≥ 18	F/M	7.7	GI GL	Dietary record	Colorectal cancer	103,020	7.7	8	Age-sex-BMI-education-PA-smoking-occupation status-energy-income-alcohol-meat-grain-menopausal status
Watling et al.,2023 [22]	UK	37–73	F/M	9.4	CHO	24 H-dietary recall	Colorectal cancer	114,217	1193	7	Occupation status-BMI-alcohol-diabetes status-energy-menopausal hormone therapy-NSAIDs-meat

Systematic review

Of the 14 studies examining the association between carbohydrate intake and colorectal cancer, 3 reported an inverse association [22, 43, 48], one study reported a positive association [45], while the remaining studies showed no significant association. 10 studies examined the association between carbohydrate intake and colon cancer. At the same time, 9 studies also investigated its relationship with rectal cancer. Among these, a negative association was observed in one study for colon cancer [50] and another for rectal cancer [41], whereas the other studies found no notable relationships for either cancer type. Three studies found a positive association between high-GI foods and increased colorectal cancer risk [45, 53, 57], although no significant associations were observed for colon or rectal cancer. Regarding the association between GL and colorectal cancer, one study indicated a positive association [45], while another identified an inverse association [50]. Additionally, one publication showed an inverse relationship with colon cancer [56], whereas no meaningful association was detected for rectal cancer.

Meta-analysis

Due to the multiple associations in this study, we divide this section into three parts based on exposure and report its association with colorectal, colon, and rectal cancers for each exposure.

Dietary carbohydrate and colorectal cancer risk

There was no significant association in the analysis of 17 effect sizes derived from 14 publications that compared the highest with the lowest carbohydrate intake categories (RR: 0.96; 95% CI: 0.89–1.03; P-value = 0.29). However, high heterogeneity among the studies led to the conduct of subgroup analyses to identify its source (I² = 55.9%; P < 0.001), and none of the subgroups were found to be the source of heterogeneity. A significant inverse association with colorectal cancer was found in subgroups of studies conducted in non-US countries, those including both men and women in the study population, and those with a follow-up duration of less than 10 years (Table 2). Sensitivity analysis confirmed that no single study significantly influenced the overall result. According to Egger’s test, no evidence of publication bias was found (P = 0.74). Eight studies were included in the linear dose-response analysis, which showed no significant association (RR: 0.99; 95% CI: 0.91–1.01; P-value = 0.08) (Table 5). No indication of a nonlinear dose-response relationship was found (P-nonlinearity = 0.92) (Fig. 2).

Table 2.

Summary risk estimate for the association between dietary carbohydrate intake and colorectal, colon, and rectal cancer

	Effect size, n	Pooled ES (95% CI)	P-value	I2(%)	p-between
Carbohydrate and colorectal cancer
Overall	17	0.96 (0.89–1.03)	0.29	55.9	0.29
Study location					0.91
US	9	0.96 (0.86–1.07)	0.44	65.7
Other country	8	0.96 (0.94–0.99)	< 0.001	45.8
Follow-up duration (year)					0.15
≥ 10	7	0.98 (0.95–1.01)	0.10	37
≤ 10	10	0.94 (0.90–0.98)	< 0.001	63.7
Sex					0.80
Female participants	9	0.94 (0.84–1.05)	0.27	48.3
Male participants	4	0.93 (0.80–1.08)	0.34	79.4
Both	4	0.96 (0.94–0.99)	< 0.001	48.2
Carbohydrate and colon cancer
Overall	12	0.99 (0.89–1.11)	0.90	33	0.12
Study location					0.87
US	8	1.01 (0.93–1.10)	0.83	47.3
Other country	4	1.03 (0.85–1.24)	0.74	3.2
Follow-up duration (year)					0.06
≥ 10	6	1.06 (0.97–1.17)	0.20	0
≤ 10	6	0.91 (0.79–1.05)	0.18	48.7
Sex					0.04
Female participants	7	0.88 (0.77–1.01)	0.06	37.5
Male participants	3	1.10 (0.92–1.32)	0.28	0
Both	2	1.08 (0.96–1.21)	0.19	0
Carbohydrate and rectal cancer
Overall	11	0.99 (0.84–1.16)	0.87	21.6	0.23
Study location					0.69
US	7	0.99 (0.84–1.15)	0.86	48.3
Other country	4	1.05 (0.81–1.36)	0.72	0
Follow-up duration (year)					0.11
≥ 10	6	1.08 (0.92–1.26)	0.36	0
≤ 10	5	0.85 (0.67–1.08)	0.18	49.3
Sex					0.36
Female participants	6	0.96 (0.77–1.21)	0.75	0
Male participants	3	0.85 (0.62–1.17)	0.31	77.7
Both	2	1.10 (0.90–1.34)	0.34	0

Values in bold indicate statistical significance (p < 0.05)

Table 5.

Summary risk estimates from linear dose-response analysis for the association of dietary carbohydrates, glycemic index, and glycemic load with colorectal, colon, and rectal cancer risk

	Publication	Pooled ES (95% CI)	P-value	I2(%)	p-heterogeneity
CHO and CRC (per 60 g)	10	0.96 (0.91–1.01)	0.08	18.1	0.27
CHO and colon cancer (per 60 g)	9	0.94 (0.88–1.01)	0.10	58	0.01
CHO and rectal cancer (per 60 g)	9	0.97 (0.92–1.02)	0.20	10.9	0.34
GI and CRC (per 5 units)	10	1.01 (0.97–1.05)	0.69	33.4	0.14
GI and colon cancer (per 5 units)	6	1.00 (0.94–1.07)	0.96	32.3	0.19
GI and rectal cancer (per 5 units)	7	1.01 (0.94–1.08)	0.86	0	0.42
GL and CRC (per 20 units)	11	1.01 (0.98–1.04)	0.44	35.5	0.11
GL and colon cancer (per 20 units)	7	0.99 (0.96–1.02)	0.42	0	0.63
GL and rectal cancer (per 20 units)	8	1.01 (0.97–1.05)	0.62	0	0.46

Fig. 2.

Non-linear dose-response association between carbohydrate, glycemic index, and glycemic load with colorectal, colon, and rectal cancer

Dietary carbohydrate and colon cancer risk

Comparison of the highest versus lowest carbohydrate intake showed no significant association between dietary carbohydrate consumption and colon cancer risk (RR: 0.99; 95% CI: 0.89–1.11; P-value = 0.90). Heterogeneity between studies was insignificant (I² = 33%; P = 0.12). Subgroup analysis revealed no significant associations based on sex (men, women, or both), geographic location (United States vs. non-United States), or follow-up duration (greater or less than 10 years) (Table 2). No individual study influenced the overall risk estimate, as confirmed by sensitivity analysis. Symmetry in the funnel plot and Egger’s test (p = 0.95) indicated no evidence of publication bias. The dose-response analyses did not show any significant linear (RR: 0.94; 95% CI: 0.88–1.01; P-value = 0.10) (Table 5) or nonlinear associations (P-nonlinearity = 0.52) between dietary carbohydrate intake and colon cancer (Fig. 2).

Dietary carbohydrate and rectal cancer risk

Dietary carbohydrate intake was not significantly associated with rectal cancer (RR: 0.99; 95% CI: 0.84–1.16; P-value = 0.87), and no significant heterogeneity was found between studies (I² = 21.6%; P = 0.23). The subgroup analysis revealed no significant associations across any subgroup. Also, no significant between-study heterogeneity was seen (I² = 21.6%; P = 0.23) (Table 2). No individual study influenced the overall result, as confirmed by the sensitivity analysis. The symmetry in the funnel plot and the results of Egger’s regression test (p = 0.07) both support the absence of publication bias. Nine eligible publications were included in the linear dose-response analysis, which found no significant association (RR: 0.97; 95% CI: 0.92–1.02; P-value = 0.34) (Table 5). Furthermore, no significant nonlinear association was seen. (P-nonlinearity = 0.11) (Fig. 2).

Glycemic index and colorectal cancer risk

A marginal positive association with colorectal cancer was observed when comparing the highest to the lowest GI categories (RR: 1.08; 95% CI: 1.00–1.16; P-value = 0.04) (Table 3). While Stata rounded the results, a more detailed examination of the values to three decimal places revealed a lower bound of 1.001, confirming that the confidence interval does not encompass one, thereby supporting the statistical significance of this association. The heterogeneity among studies was found to be non-significant (I² = 33%; P = 0.91). Subgroup analysis identified positive significant associations in specific subgroups. Studies conducted in the United States demonstrated a 12% increase in colorectal cancer risk associated with a higher GI. Also, direct associations were observed in studies with a follow-up duration of less than 10 years and those focusing on male populations. Sensitivity analysis confirmed that no individual study impacted the overall result. Publication bias was evaluated by examining the symmetry of the funnel plot and conducting Egger’s test (p = 0.95), and no evidence of publication bias was observed. No linear association was found in the dose-response analysis (RR: 1.01; 95% CI: 0.97–1.05; P-value = 0.69) (Table 5). Additionally, there was no significant nonlinear dose-response association (P-nonlinearity = 0.81) (Fig. 2).

Table 3.

Summary risk estimate for the association between glycemic index and colorectal, colon, and rectal cancer

	Effect size, n	Pooled ES (95% CI)	P-value	I2(%)	p-between
GI and colorectal cancer
Overall	17	1.08 (1-1.16)	0.04	33	0.91
Study location					0.13
US	9	1.12 (1.05–1.20)	< 0.001	32.6
Other country	8	1.02 (0.92–1.13)	0.65	28.8
Follow-up duration (year)					0.26
≥ 10	11	1.05 (0.97–1.14)	0.22	15.8
≤ 10	6	1.12 (1.04–1.21)	< 0.001	53.7
Sex					0.72
Female participants	10	1.07 (0.99–1.15)	0.10	18.6
Male participants	3	1.11 (1.01–1.22)	0.02	40.6
Both	4	1.12 (0.95–1.33)	0.17	66.2
GI and colon cancer
Overall	10	1.04 (0.94–1.16)	0.44	17	0.28
Study location					0.38
US	3	1.09 (0.95–1.26)	0.21	0
Other country	7	1.01 (0.89–1.14)	0.94	39.9
Follow-up duration (year)					0.61
≥ 10	8	1.03 (0.92–1.15)	0.60	33.6
≤ 10	2	1.09 (0.90–1.31)	0.37	0
Sex					0.91
Female participants	6	1.06 (0.94–1.19)	0.84	0
Male participants	2	1.02 (0.83–1.25)	0.33	0
Both	2	1.01 (0.95–1.15)	0.90	86.6
GI and rectal cancer
Overall	10	1.04 (0.89–1.21)	0.64	0	0.52
Study location					0.32
US	3	1.19 (0.87–1.64)	0.27	0
Other country	7	0.99 (0.83–1.19)	0.93	14.5
Follow-up duration (year)					0.42
≥ 10	8	1.01 (0.85–1.20)	0.92	5.1
≤ 10	2	1.20 (0.81–1.79)	0.35	0
Sex					0.87
Female participants	6	1.03 (0.83–1.28)	0.79	23
Male participants	2	0.99 (0.72–1.36)	0.93	0
Both	2	1.11 (0.80–1.54)	0.52	0

Values in bold indicate statistical significance (p < 0.05)

Glycemic index and colon cancer risk

A total of 10 effect sizes from 7 studies were analyzed to examine the association between GI and colon cancer risk. The comparison between the highest and lowest GI categories did not indicate a marginally significant association (RR: 1.04; 95% CI: 0.94–1.16; P-value = 0.44) (Table 3). Between-study heterogeneity was not substantial (I² = 17%; P = 0.28). The sensitivity analysis showed no study significantly impacted the overall result. The funnel plot appeared symmetrical, with Egger’s test further supporting the lack of publication bias (P = 0.39). The analysis of 6 effect sizes in the linear dose-response model showed no significant association(RR: 1.00; 95% CI: 0.94–1.07; P-value = 0.96) (Table 5). Additionally, no evidence of a nonlinear connection was found(P-nonlinearity = 0.05) (Fig. 2).

Glycemic index and rectal cancer risk

We included 10 effect sizes from 7 publications to analyze the association between GI and rectal cancer. By comparing the highest versus the lowest categories, we found no significant association (RR: 1.04; 95% CI: 0.89–1.21; P-value = 0.64) (Table 3). There was no evidence of significant between-study heterogeneity (I² = 0%; P = 0.52). Subgroup analyses were conducted, but none of the subgroups showed a significant association. Sensitivity analysis showed that the final RR was unaffected by any individual study. No asymmetry was seen in the funnel plot, and Egger’s test robustly confirmed the absence of publication bias (p = 0.82). A dose-response examination was carried out, and no significant linear relationship was found (Table 5). Furthermore, no nonlinear association was found (P-nonlinearity = 0.62) (Fig. 2).

Glycemic load and colorectal cancer risk

We included 20 effect sizes from 15 studies, comparing the highest with the lowest GL categories to assess their potential association with colorectal cancer, but no significant association was found (RR: 0.99; 95% CI: 0.91–1.08; P-value = 0.84), and between-study heterogeneity was also not significant. (I² = 33.6%; P = 0.72). The subgroup analysis did not find any significant associations in the different subgroups (Table 4). Additionally, Sensitivity analysis indicated that the overall relative risk (RR) was not influenced by any single study. No asymmetry was detected in the funnel plot, and Egger’s test further confirmed the absence of publication bias (P = 0.07). No significant association was observed in the linear dose-response analysis (RR: 1.01; 95% CI: 0.98–1.04; P-value = 0.44) (Table 5), nor was any significant association found during the examination of the nonlinear dose-response model (P-nonlinearity = 0.16) (Fig. 2).

Table 4.

Summary risk estimate for the association between glycemic load and colorectal, colon, and rectal cancer

	Effect size, n	Pooled ES (95% CI)	P-value	I2(%)	p-between
GL and colorectal cancer
Overall	20	0.99 (0.91–1.08)	0.84	33.6	0.72
Study location					0.94
US	11	0.98 (0.91–1.07)	0.69	46.4
Other country	9	0.98 (0.87–1.10)	0.70	19.6
Follow-up duration (year)					0.28
≥ 10	12	1.02 (0.93–1.11)	0.73	25.9
≤ 10	8	0.94 (0.86–1.04)	0.24	44.6
Sex					0.93
Female participants	12	0.97 (0.89–1.06)	0.51	25.7
Male participants	4	0.99 (0.88–1.13)	0.93	63.2
Both	4	1.01 (0.81–1.25)	0.94	45.8
GL and colon cancer
Overall	13	0.98 (0.85–1.12)	0.77	56	0.77
Study location					0.19
US	5	1.05 (0.94–1.18)	0.35	63.2
Other country	8	0.94 (0.82–1.07)	0.35	52.3
Follow-up duration (year)					0.42
≥ 10	9	1.03(0.93–1.14)	0.60	63.8
≤ 10	4	0.95 (0.81–1.12)	0.53	33.9
Sex					0.03
Female participants	8	0.94 (0.83–1.06)	0.28	0
Male participants	3	1.16 (1.00-1.33)	0.04	67.7
Both	2	0.85 (0.65–1.12)	0.25	89.2
GL and rectal cancer
Overall	13	1.05 (0.90–1.24)	0.52	3.5	0.41
Study location
US	5	1.05 (0.81–1.35)	0.72	41.2	0.95
Other country	8	1.06 (0.87–1.29)	0.58	0
Follow-up duration (year)					0.69
≥ 10	9	1.08 (0.89–1.31)	0.45	0
≤ 10	4	1.01 (0.78–1.31)	0.93	36.5
Sex					0.59
Female participants	8	1.01 (0.82–1.23)	0.94	22.4
Male participants	3	1.05 (0.77–1.44)	0.75	0.30
Both	2	1.28 (0.84–1.95)	0.24	0

Glycemic load and colon cancer risk

There was no significant association between GL and colon cancer in the analysis of 13 effect sizes from 9 studies comparing the highest with the lowest categories (RR: 0.98; 95% CI: 0.85–1.12; P-value = 0.77) (Table 4). However, significant heterogeneity between studies was observed (I² = 56%; P < 0.001), with subgroup analysis identifying sex as the main contributor to heterogeneity. Nevertheless, the association between GL and colon cancer remained non-significant across all subgroups. According to the sensitivity analysis, no individual study influenced the overall effect size. Furthermore, symmetry was observed in the funnel plot, and Egger’s test also indicated no evidence of publication bias(p = 0.46). The linear dose-response analysis demonstrated no significant association (RR: 0.99; 95% CI: 0.96–1.02; P-value = 0.42) (Table 5), and similarly, the nonlinear dose-response analysis revealed no significant association (P-nonlinearity = 0.68) (Fig. 2).

Glycemic load and rectal cancer risk

No significant association between GL and rectal cancer was observed when comparing the highest to the lowest category (RR: 1.05; 95% CI: 0.90–1.24; P-value = 0.52) (Table 4). Moreover, heterogeneity across studies was not significant (I2 = 3.5%; P = 0.41), and subgroup analyses did not reveal any substantial association (Table 4). The symmetry of the funnel plot, along with the results from Egger’s test, indicated the absence of publication bias (p = 0.87). According to the sensitivity analysis, the results were not driven by any single study. In the linear dose-response analysis, which included 8 studies, no significant association was observed. (RR: 1.01; 95% CI: 0.97–1.05; P-value = 0.62) (Table 5). Additionally, no nonlinear dose-response association was found between GL and rectal cancer (P-nonlinearity = 0.55) (Fig. 2).

Discussion

We systematically and meta-analytically reviewed previous studies on linking dietary carbohydrate intake, GI, GL, and colorectal cancer risk. Our investigation found no significant association between dietary carbohydrate intake and the risk of colorectal, colon, or rectal cancers in the comparison of the highest and lowest intake groups. However, subgroup analyses revealed an inverse association between carbohydrate consumption and colorectal cancer risk in specific subgroups, including studies with follow-up periods shorter than 10 years, populations of both sexes, and those conducted in non-US countries.

Diversity in dietary carbohydrate patterns across regions may explain this inverse association observed in subgroup analyses. More than half of the studies included in our meta-analysis were conducted in the United States, where carbohydrate intake comes from a wide variety of sources with both high and low glycemic indices (such as whole grain, potatoes, and pasta) [60, 61]. This high variability in carbohydrate sources and glycemic quality in this country may have attenuated or masked any real association between total carbohydrate intake and colorectal cancer risk, which could explain why no significant association was found in the overall analysis. Additionally, certain carbohydrate-rich foods are believed to possess cancer-preventive properties, potentially mitigating negative effects and resulting in opposing outcomes [62, 63]. Although the American Institute for Cancer Research & World Cancer Research Fund indicate that diets rich in dietary fiber from sources such as fruits, vegetables, legumes, and whole grains have been strongly associated with a reduced risk of colorectal cancer [64], it is possible that the protective effects observed in previous studies may be attributed primarily to the fiber content of these carbohydrate-rich foods [65]. We found no notable links between carbohydrate and colorectal cancer risk. In line with our results, Makarem et al. demonstrated that there is no association between carbohydrate intake and CRC in the Framingham offspring cohort [54]. They profess that the relatively modest strength of links between dietary components and cancer risk poses challenges for detection, particularly in studies with small sample sizes or few colorectal cancer events within the cohort [54]. Moreover, the results are broadly consistent with those of two large prospective cohort studies of Canadian [44] and US women [46], which have follow-ups of up to 20 years. Nevertheless, this limitation is inherent to observational study designs, which can not establish causality and may therefore contribute to the modest effect sizes observed.

Elevated GI was associated with a marginally significant increase in the risk of colorectal cancer. This relationship persisted in subgroup analyses, particularly among men, studies conducted in the United States, and those with follow-up periods of less than 10 years. Analysis revealed no significant link between GI and the risk of colon or rectal cancer. Similarly, GL was not linked to the risks of colorectal, colon, or rectal cancer. Moreover, both linear and nonlinear dose-response analyses revealed no significant relationship between consumption of carbohydrates, GI, GL, and risk of colorectal cancer.

Previous investigations on this issue have produced mixed, mainly null findings. A 2019 study found similar results. This study studied carbs, GI, and GL in colon cancer and other cancers. The highest and lowest GI groups showed that a higher GI increased colorectal cancer risk. No connection was found between other exposures and colorectal cancer, and colon and rectal cancer studies were not done [21]. Our research and the meta-analyses reveal a link between greater GI and colorectal cancer risk. IGF-1 elevation may explain many of the mechanisms of this association. High insulin levels promote IGF-1 release, which helps peripheral tissues absorb glucose. IGF-1 increases insulin sensitivity, regulating blood glucose levels under normal physiological settings [66]. Elevated insulin secretion, triggered by an increased GI, results in a decrease in IGF-binding proteins, thereby augmenting the levels of bioavailable IGF. As a potent mitogen, IGF facilitates the proliferation of malignant colonocytes while concurrently inhibiting apoptotic processes, contributing to the progression of colorectal cancer [67, 68]. IGF-1’s indirect actions on colonocytes may also contribute to colorectal cancer development. In CRC cells, IGF-1 increases glucose absorption, shifting metabolism to glycolysis. Cancer cells prefer glycolysis for ATP synthesis, even with oxygen, due to the Warburg effect. This aerobic glycolysis increases lactate generation and glucose metabolism genes like GLUT1. As glycolysis dominates energy generation, lactate synthesis sustains anabolic activities needed for fast cellular proliferation, supporting tumor development [68]. Furthermore, IGF-1 is essential in modulating various intracellular signaling cascades, with the PI3K/Akt/mTOR axis being one of the most critical. Upon activation by IGF-1, this axis orchestrates the expression of key metabolic genes essential for tumor progression. Notably, IGF-1 has been shown to upregulate HIF-1α, a central regulator of the Warburg effect and a fundamental driver of tumorigenesis. The subsequent activation of HIF-1α fosters the activation of genes that contribute to both angiogenesis and metabolic reprogramming, such as VEGF, thereby promoting the adaptive metabolic changes required for sustained tumor growth [68–70].

A 2017 meta-analysis examining the association between dietary carbohydrates and colorectal cancer in observational studies found no significant association. Similarly, subgroup analyses limited to cohort studies failed to identify a meaningful association. However, our study contrasts these findings, revealing a significant inverse association in certain subgroups [20]. This variance may be due to dietary differences, such as carbohydrate kinds, fiber consumption, or protective effects. Diets with a high content of refined carbohydrates and a low amount of fiber promote gut microbiota inflammation and dysbiosis, which are linked to colorectal cancer. Whole grains, legumes, and fruits, rich in dietary fiber, antioxidants, and other bioactive substances, may protect against colorectal carcinogenesis. Increasing gut microbial diversity, insulin sensitivity, and reducing oxidative stress and inflammation may provide these protective benefits [65, 71, 72]. Also, Food frequency surveys may not accurately reflect the relationship between nutrient consumption and illness due to measurement errors [59, 73, 74]. As with prior research, our findings may have been influenced by inaccurate recollection of carbohydrate and sugar consumption, changes in diet over time, or underreporting of diet.

Several features strengthen this meta-analysis’s validity and reliability. First, the large sample size, large number of included studies, and inclusion of the latest research strengthened our analysis, revealing the relationship between exposures and outcomes. Second, we employed prospective cohort studies to reduce memory and selection biases in case-control designs for more accurate findings. Isolating colon and rectal cancers as independent endpoints allows for a more sophisticated examination of exposure-type relationships, the study’s third strength. The colon and rectal cancer subgroup analyses added specificity and robustness to our results. Fourth, a dose-response study examined linear and non-linear connections to better understand exposure-outcome relationships. Fifth, this study examined the link between dietary carbs, GI, and GL, unlike earlier colorectal cancer studies that focused on one or a few of these exposures. Finally, meta-analyses may have publication bias. Visual examination of funnel plots and statistical evaluations like Egger tests showed no substantial publication bias, which mitigates its possible influence on our results. This study is not without its limitations, several of which are intrinsic to the nature of observational studies and meta-analytic approaches. Residual confounding may arise from unmeasured or insufficiently controlled variables, potentially undermining the validity and accuracy of the results in the studies included in this analysis. Cohort studies predominantly employed FFQs for dietary assessment; however, the FFQs varied in terms of the number of items included. Furthermore, some studies utilized alternative dietary assessment methods, such as dietary recalls or food records. These methodological inconsistencies across studies can introduce potential biases, including the underreporting of dietary carbohydrate intake. Certain studies were excluded from the dose-response analysis because they lacked sufficient data for inclusion. Lastly, the observed inverse association between carbohydrate intake and colorectal cancer risk in certain subgroups may be attributed to the limited number of studies. This finding requires further exploration, and additional research in this area could provide more clarity and a more robust understanding of the potential association.

Conclusion

A marginally positive association was identified between GI and colorectal cancer risk. Subgroup analyses revealed an inverse association between carbohydrate intake and colorectal cancer in specific subgroups, including men, women, studies conducted in non-US populations, and cohorts with follow-up durations shorter than 10 years. However, no significant associations were observed for other exposures and outcomes. The dose-response analysis did not indicate any significant linear or nonlinear associations.

Abbreviations

GI

glycemic index

GL

glycemic load

CRC

colorectal cancer

CRP

C-Reactive Protein

Authors’ contributions

MMM and NAH were responsible for the literature search, data extraction, study conception, and manuscript drafting. MMM and MZ contributed to the study conception and handled the data analysis, while LA provided supervision throughout the process and approved the final manuscript. All authors assume complete responsibility for the analysis, interpretation, and presentation of the findings in this report.

Funding

This research received financial support from the Tehran University of Medical Sciences (TUMS).

Data availability

The data underlying the findings of this study can be obtained from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

The study is deemed exempt from receiving ethical approval.

Consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.