Causal associations among albumin, C-reactive protein, body mass index, and the risk of colorectal cancer

Abstract

Purpose

A Mendelian randomization model analysis was employed to investigate the causal direction between CRP (C-reactive protein), Alb(albumin), and BMI(body mass index), as well as their impact on colorectal cancer (CRC) risk and postoperative prognosis.

Methods

Two-sample Mendelian randomization (MR) was conducted, with BMI set as the exposure variable and CRP, Alb, and CRC as the outcome variables, to determine if CRP and Alb mediate the causal relationship between BMI and CRC. Additionally, this relationship was examined using individual-level data from a retrospective cohort analysis of local cases, which included 374 incident cases of CRC.

Results

The analysis of the mediating effect of Alb on BMI and CRC risk demonstrated that in European and East Asian populations, BMI had a promoting effect on CRC risk (OR: 1.1903 and 1.3243, p < 0.05). A positive causal relationship was observed between BMI and Alb, but the effect was opposite (OR: 0.8671 and 0.9283, p < 0.05). Alb exhibited a protective effect against CRC (OR: 0.8707 and 0.7157, p < 0.05). The mediating effect of CRP indicated that BMI in European and East Asian populations was associated with an increased risk of CRC (OR: 1.0013 and 1.3243, p < 0.05). BMI and CRP had a positive causal relationship, with consistent effects (OR: 1.4306 and 1.1008, p < 0.05). CRP had a weak, albeit unclear, protective effect on CRC risk (OR: European, 0.9985; p < 0.05, and East Asian, 0.9992; p > 0.05). Relative to the MR results, a high BMI and Alb, along with a low CRP, were observed to be associated with more prolonged overall survival (OS) in the retrospective cohort univariate analysis (HR: 0.89, 0.92, and 1.02, P < 0.05). After adjusting for Alb, the association between BMI and CRC overall survival remained significant (HR: 0.91, P < 0.05). However, upon adjustment for CRP, BMI was not found to be independently associated with overall survival in patients with CRC.

Conclusions

The findings of this study support that BMI and related factors have a causal association with the risk of colorectal cancer.

Supplementary Information

The online version contains supplementary material available at 10.1007/s12672-025-04334-w.

Introduction

Colorectal cancer is one of the most common cancers in the world, representing about 10% of all cancer cases, and is the second leading cause of cancer-related deaths worldwide [1, 2]. It predominantly affects older adults, with most cases occurring in individuals aged 50 and above [2]. Among lifestyle factors, being overweight increases the risk of developing and dying from colorectal cancer. Additionally, physical inactivity, a diet high in red and processed meats, cooking meat at very high temperatures, alcohol consumption, and smoking are all associated with a higher risk of colon cancer and can negatively impact prognosis [3].

Albumin (Alb) is a protein synthesized by the liver that aids in maintaining blood volume, transports hormones, vitamins, and enzymes throughout the body, and prevents fluid from leaking from blood vessels [4]. CRP is a protein that the liver produces in response to inflammation [5]. Elevated CRP levels are linked to an increased risk of heart disease and other inflammatory conditions [5, 6]. BMI is a measure of body fat, BMI = weight (kg)/(height (m)) ² [7, 8].

A strong positive correlation exists between baseline CRP concentrations and BMI, with CRP levels increasing as BMI increases [5, 6, 9]. Losing weight has been shown to lower CRP levels [10]. CRP and Alb respond to inflammation in the body and are regulated by the IL-6 inflammatory pathway [11]. Previous studies by our research group and others have proven that the CRP/Alb ratio is closely related to inflammatory bowel disease and colorectal cancer [6, 12, 13]. Additionally, within BMI categories, a CRP/Alb ratio of > = 0.097 was a better predictor of polycystic ovary syndrome (PCOS) than insulin resistance (IR) and the free androgen index [14]. Nutritional status may affect Alb levels, suggesting a correlation with BMI [15]. Studies have examined the relationship between BMI, Alb as an inflammation biomarker, and colorectal cancer risk across various ethnic groups [16]. Each standard deviation increase in Alb was associated with a decreased risk of colorectal cancer. Concurrent mediation analysis indicated that BMI did not mediate the relationship between Alb and CRC [16]. Alb and BMI are both associated with colorectal cancer risk, but they seem to exert independent effects [17]. Recent studies have emphasized the prognostic value of systemic inflammatory and nutritional biomarkers in colorectal cancer. Combined inflammatory indicators such as monocyte count, neutrophil-based indices, and the albumin-to-globulin ratio have been shown to predict survival outcomes in CRC patients [18, 19]. In addition, several inflammation- and nutrition-based scoring systems have been developed or validated for stage I–III CRC, supporting the relevance of biomarkers such as CRP and Alb in prognostic assessment [20, 21]. Furthermore, biochemical parameters, including serum calcium, have also been shown to have prognostic significance in CRC [22]. These findings highlight the importance of inflammatory and nutritional pathways in CRC progression and provide a rationale for examining CRP and Alb in the present study.

Mendelian randomization (MR) is a method used in epidemiology to investigate the causal effect of an exposure on an outcome [23]. Mendelian randomization relies on the random assignment of genetic variation from parents to offspring, forming the basis of this method [24]. This design reduces reverse causation and confounding factors. Therefore, MR has become a reliable research method for solving complex problems. Although studies have explored the role of CRP in BMI and prostate cancer risk, the results remain inconclusive [25]. The results suggest that CRP may mediate this relationship; however, further research is still needed to confirm whether a similar mechanism exists in colorectal cancer. While CRP levels can indicate inflammation, they do not show the cause of the inflammation. Therefore, high CRP levels may be due to factors other than cancer or high BMI. This study used CRP and Alb to proxy for human chronic inflammation. To analyze and explore the impact of CRP and Alb on BMI and the risk and prognosis of colorectal cancer.

Methods and materials

Study design

Two-sample MR was performed, with BMI designated as the exposure variable and CRP, Alb, and CRC as the outcome variables, as established in previous research. Additionally, BMI, CRP, and Alb were utilized as exposure variables, with CRC as the outcome variable, to investigate whether CRP and Alb mediate the causal relationship between BMI and CRC. Multivariate Mendelian randomization (MVMR) studies were designed to quantify the proportion of the effect mediated by the identified mediators. The BMI, CRP, and Alb data of 477 patients with stage I-III postoperative colorectal cancer admitted to the Gastrointestinal Surgery Department of Guangxi Medical University Cancer Hospital and Chenzhou No. 1 People’s Hospital from October 2013 to September 2019 were retrospectively analyzed. All patients received abdominal B-ultrasound, magnetic resonance imaging, and computed tomography, along with assessments for relevant clinical symptoms and physical signs. The inclusion criteria were: (1) colorectal cancer (CRC) diagnosed by histopathology or radiological findings, and (2) receipt of surgical R0/R1 resection or chemotherapy with leucovorin/fluorouracil/oxaliplatin (FOLFOX) or CapOX regimens. Patients were excluded if they: (1) showed clinical signs of infection or other inflammatory conditions, (2) had undergone anticancer treatment before laboratory evaluation, or (3) had primary malignancies originating in other organs. Of these, 374 cases were included in this study, while 103 were excluded due to incomplete medical information or a lack of follow-up, as detailed in Table 1; Fig. 1. The overall survival (OS) in this study is defined as the time interval between the patient’s enrollment date and the date of death from any cause. Patients underwent complete follow-up after receiving initial treatment. The follow-up period began on the date of first diagnosis of colorectal cancer (CRC) and ended either in September 2019 or at the time of death.

Table 1.

Descriptive statistics of local retrospective study cohort

Characteristic	Alive, N = 289	Dead, N = 85	p-value1
Height, cm			0.948
Median (IQR)	163.00 (158.00, 169.00)	163.00 (156.00, 168.00)
Weight, kg			0.019
Median (IQR)	58.00 (50.00, 65.00)	54.00 (48.00, 65.00)
BMI			0.003
Median (IQR)	21.79 (19.72, 23.89)	20.70 (18.67, 23.38)
WBC,10^9/L			0.009
Median (IQR)	6.50 (5.34, 7.64)	7.06 (6.15, 8.18)
PLT,10^9/L			0.547
Median (IQR)	287.00 (240.45, 355.15)	277.12 (239.98, 372.84)
NEU%			0.008
Median (IQR)	59.30 (52.90, 66.10)	63.20 (56.30, 68.70)
NEU#,10^9/L			0.009
Median (IQR)	3.80 (2.95, 4.92)	4.30 (3.63, 5.34)
MONO%			0.407
Median (IQR)	6.90 (5.60, 8.10)	7.10 (6.00, 8.20)
MONO#,10^9/L			0.002
Median (IQR)	0.44 (0.36, 0.55)	0.51 (0.40, 0.68)
LYM%			0.002
Median (IQR)	28.50 (22.70, 35.10)	25.30 (19.40, 31.90)
LYM#,10^9/L			0.795
Median (IQR)	1.82 (1.44, 2.16)	1.81 (1.32, 2.16)
EO%			0.826
Median (IQR)	3.20 (1.90, 5.40)	2.80 (1.50, 4.50)
EO#,10^9/L			0.804
Median (IQR)	0.20 (0.13, 0.34)	0.20 (0.11, 0.37)
BASO%			0.013
Median (IQR)	0.50 (0.40, 0.70)	0.40 (0.30, 0.60)
BASO#,10^9/L			0.618
Median (IQR)	0.03 (0.02, 0.05)	0.03 (0.02, 0.05)
LMR			0.003
Median (IQR)	0.47 (0.34, 0.66)	0.41 (0.28, 0.54)
CRP,mg/L			< 0.001
Median (IQR)	3.57 (1.56, 4.99)	17.60 (4.17, 50.50)
Alb,g/L			0.016
Median (IQR)	39.90 (37.00, 42.70)	37.90 (33.30, 41.20)
TIME, month			< 0.001
Median (IQR)	47.00 (30.00, 59.00)	23.00 (14.00, 37.00)
Age, year			0.018
Median (IQR)	55.00 (47.00, 64.00)	59.00 (50.00, 68.00)
CEA,ng/mL			0.004
Median (IQR)	3.33 (1.86, 8.08)	12.64 (5.42, 50.79)
CA199, U/mL			0.007
Median (IQR)	11.68 (6.40, 27.43)	17.60 (5.94, 62.68)

¹Welch Two Sample t-test

Fig. 1.

Study design. We triangulated findings from two study designs: a Mendelian randomisation analysis and Retrospective analysis to estimate the causal effect of BMI, Alb, and CRP on CRC

The GWAS dataset used in this section is publicly available and approved by the relevant ethical review committee, so no additional ethical clearance was required. The use of public data and clinical data collection adheres to the guidelines of the Declaration of Helsinki. Clinical data collection is supervised and approved by the Ethics Committee of Guangxi Medical University Cancer Hospital (KY-2022-490).

Data source

The GWAS summary data from the IEU GWAS data portal were utilized to perform the two-sample MR analysis (https://gwas.mrcieu.ac.uk/) [26]. Information on instrumental variables for genetic variants associated with BMI, Alb, and CRC was obtained from the IEU analysis via the EBI database, which contains comprehensive GWAS summary data for the European population. BMI data included 457,756 samples. The Alb data included 315,268 samples [27, 28]. Data on outcome variables for CRC included 19,948 cases of CRC and 12,124 controls [29]. The validation set was extracted separately from Biobank Japan’s release of disease traits (East Asia). BMI data included 158,284 samples [30]. The Alb data included 102,223 samples [31]. Data on outcome variables for CRC included 6,692 cases of CRC and 27,178 controls [27]. Moreover, the instrumental variables for BMI, CRP, and CRC were extracted separately from datasets ieu-b-40, ieu-b-35, and ukb-d-C18(colon cancer only, not CRC) [26, 33, 34]. The exposure and outcome variables of the validation set were extracted from Biobank Japan’s release of disease traits, including bbj-a-1 [30], bbj-a-14 [31], and bbj-a-76 [32], as shown in Table 2.

Table 2.

Summary of the GWAS included in this MR study

Exposures/outcomes		GWAS ID	Consortium	Ethnicity	Sample sizes	Number of SNPs	Sex	Year
Body mass index		GCST90025994	EBI	European	457,756	9,800,000	Male and Female	2021
Serum albumin		GCST90018945	EBI	European	315,268	19,053,186	Male and Female	2021
Malignant neoplasm of colorectal		GCST012879	EBI	European	32,072 (ncase: 19,948; ncontrol: 12,124)	38,356,021	Male and Female	2018
Body mass index		bbj-a-1	BBJ	East Asia	158,284	5,961,600	Male and Female	2018
Serum albumin		bbj-a-9	BBJ	East Asia	102,223	6,108,953	Male and Female	2018
Malignant neoplasm of colorectal		bbj-a-76	BBJ	East Asia	33,870 (ncase: 6,692; ncontrol:27,178)	7,492,477	Male and Female	2019
Body mass index		ieu-b-40	MR-Base	European	461,460	9,851,867	Male and Female	2018
C-reactive protein		ieu-b-35	MR-Base	European	204,402	2,414,379	Male and Female	2018
Malignant neoplasm of colorectal		ukb-d-C18	UKB	European	361,194 (ncase: 2,226; ncontrol: 358,968)	10,833,390	Male and Female	2018
Body mass index		bbj-a-1	BBJ	East Asia	158,284	5,961,600	Male and Female	2018
C-reactive protein		bbj-a-14	BBJ	East Asia	75,391	6,108,953	Male and Female	2018
Malignant neoplasm of colorectal		bbj-a-76	BBJ	East Asia	33,870 (ncase: 6,692; ncontrol:27,178)	7,492,477	Male and Female	2019

Instrumental variables

For two-sample MR analysis, we ensure that the instrumental variants (IVs) are independent of each other, meaning there is no significant linkage disequilibrium (LD) (p < 5 × 10^{− 8}). After completing the data analysis, we performed LD Clumping on it, specifying the LD reference as “EUR” (with “EAS” used for East Asia); the threshold for r2 was set to 0.001. The operation involves interacting with the OpenGWAS API. Phenotypes associated with the remaining SNPs were identified by searching the human genotype-phenotype association database. Single-nucleotide polymorphisms (SNPs), widely recognized for their influence on outcome variables, were excluded. The outcome GWAS data were extracted from an online database using the OpenGWAS API. SNPs that fit the hypothesis were obtained from the results. (Hypothesis 1: SNPs functioning as valid instrumental variables are hypothesized to impact CRC risk exclusively through the specified exposure. 2: Assumptions involve SNPs being associated with exposures in the population. Hypothesis 3: Assumptions that SNPs do not directly affect outcomes except through exposure [35]. After extracting the exposure and outcome data sets, we merged them to ensure uniformity in their effect scales. Some SNPs are directional; their relationship to exposure or outcome cannot be determined. Eliminate these SNPs by judging the size of the effect allele frequency (EFA). SNPs that are inconsistent or cannot be matched in the exposure and outcome data. Identify and eliminate these incompatible SNPs.

Mendelian randomization analysis

MR analysis uses 6 methods by default (MR Egger, Weighted median, Inverse variance weighted (IVW), IVW random-effects model, Simple mode, Weighted mode). Odds ratios (OR) and 95% confidence intervals (CI) were provided for all methods. We did not use the robust adjusted profile scores (RAPS) method. The heterogeneity test is primarily used to assess whether the SNPs being used are heterogeneous. Heterogeneity refers to the possible differences in the relationship between exposure and outcome for different SNPs, which can affect the reliability of MR analysis. When we observed significant heterogeneity in our analysis (p < 0.05 for Cochran’s Q-test) [36], a random effects model was used, and a fixed effects model was used when heterogeneity was not significant. The horizontal pleiotropy test is used in MR to detect whether an SNP affects multiple outcomes.

We employ two methods for detection: the “MR-PRESSO global test” to identify horizontal pleiotropy and the “MR-PRESSO outlier test” to remove abnormal SNPs (outliers), providing corrected estimates [37]. The MR-Egger [38] intercept test, like a global test of MR-PRESSO, was employed to detect the presence of pleiotropy; the Egger Intercept is the intercept in the Egger method. In general, if the value is significantly different from 0, it may indicate the presence of horizontal pleiotropy.

The Steiger test [39] validates the hypothesis that the exposure caused the outcome. We use the Steiger test to calculate the variance that the instrumental variable (IV) explains for both exposure and outcome. If the variance explained by the outcome is smaller than that explained by the exposure, it confirms the correct direction.

Mediation analysis

For the mediation analyses, we applied a two-step MR framework, which relies on several key validity assumptions. First, the genetic instruments must be robustly associated with the exposure of interest (relevance assumption). Second, the instruments should not be associated with any confounders of the exposure–mediator or mediator–outcome relationships (independence assumption). Third, the instruments must influence the outcome only through the exposure and mediator, without alternative pathways, thereby satisfying the exclusion restriction and minimizing the risk of collider bias. We examine the causal effect of exposure (BMI) on the outcome (CRC) with Alb (or CRP) as mediators. Both the exposure and the mediator have their respective associated instrumental variables. Figure 1 presents a causal-directed acyclic graph illustrating the relationship between these variables. The effect of exposure on the mediator (BMI on Alb (or CRP)) and the effect of exposure on the outcome [(BMI, Alb or CRP) on CRC] on the mediating variable were assumed to be linear, with no interactions. Figure 1 does not include other confounding factors. Calculating the mediating effect involves two steps. First, the effect of exposure on the mediating variable is determined. Second, the effect of the mediating factor on the outcome is calculated. The mediating effect is then obtained by multiplying these two coefficients. The direct effect is obtained by subtracting the mediating effect from the total effect. The IV we used in the second step excludes the one used in the first step.

The detailed description is first to calculate the β_ALL value from BMI to CRC to clarify whether the mechanism of BMI on the pathogenesis of colorectal cancer can be passed. Then, the mediating effect is calculated. In the first step, the TSMR method obtains the β1 value from BMI to Alb (or CRP); in the second step, the β2 value from Alb (or CRP) to CRC is obtained. The indirect effect of the mediating effect is “β1*β2”, the direct impact is “β_ALL-β1*β2”, and the standard error of the effect size (SE) is SE = sqrt (β1² *SE2² + β2² *SE1²), where SE1 and SE2 are calculated in the first and second steps, respectively. The sqrt () function is used to find the square root of a given value. The statistic of the Sobel test equation Z = β1*β2/SE, the p-value formula is p = 2*pnorm(q = abs(Z), lower.tail = FALSE), where the function of norm() function is, in the normal distribution on the data curve of, returns the integral from negative infinity to q, where q refers to a Z value. When lower.tail = FALSE, calculate the probability that the given quantile reaches positive infinity, and the abs() function is used to find the absolute value of the given value. The 95% confidence interval calculation formula is that the lower bound is Lci = β1*β2-1.96.96*SE and the upper bound is Uci = β1*β2 + 1.96*SE. The calculation formula for the proportion of the mediating effect is β1*β2/β_ALL, the lower bound of the 95% confidence interval is Lci/βALL, and the upper bound is Uci/β_ALL.

Multivariable Mendelian randomization (MVMR) methods were employed to validate the association between other exposures and the outcome CRC, adjusting for Alb (CRP) or BMI. The β value of Alb (CRP) on CRC, after controlling for BMI, is denoted as β3. The β value of BMI on CRC, after controlling for Alb (CRP), is denoted as β4. MVMR can obtain the direct effect of exposure to BMI β5 in the mediation model. The indirect effect is calculated as β1β3, and the proportion of the mediating effect is calculated as β1β3/β1β3 + β4.

Cohort study between BMI, CRP, Alb, and CRC

We conducted a retrospective cohort analysis of three exposures: BMI, CRP, Alb, and incident colorectal cancer. BMI, CRP, and Alb are all continuous variables, and other covariates are also included: age, gender, site of onset, pathological differentiation, CEA, CA19-9, lymph node metastasis, and distant metastasis. We also collected continuous data on counts of five leukocyte subtypes. The COX regression was then used for CRC prognosis. The main retrospective analysis, including BMI, CRP, and Alb variables, is called “Model 1,2,3,4”, the minimally adjusted model. A fully adjusted analysis was performed, incorporating all additional variables, termed “Model 5,” which included all additional covariates. Subsequently, another set of additional analyses was performed, “Model 6”, adding five leukocyte subtype counts to the model to investigate potential biases due to the association of BMI, CRP, and Alb with immune cells. An analysis that studies variables such as BMI, CRP, and Alb individually is called “univariate,” while an analysis that combines them is called “multivariate.”

Statistical analysis

All MR analyses were conducted using R (version 4.2.3) and GraphPad Prism 9.0 within a Windows 11 environment. Genetic data preparation and MR analyses were performed utilizing the TwoSampleMR (version 0.5.8) package, while MVMR analyses were carried out using the ‘MVMR’ R package.

Results

Mendelian randomization of BMI, Alb, and CRP on CRC

Before MR analysis, we will estimate the average F statistic for BMI, Alb, and CRP. From a traditional statistical perspective, when the F statistic is less than 10, we typically consider the genetic variant to be a weak instrumental variable, which can lead to biased results. For the outcomes CRC and Alb, the F values for BMI were 67.64 (Europe, SNP) and 56.17 (Asia, SNP); for the outcomes CRC and CRP, the F values for BMI were 70.44 (Europe, SNP) and 56.17 (Asia, SNP).); for the outcome CRC, the F values of Alb were 82.87 (Europe, SNP) and 61.44 (Asia, SNP); the F values of CRP were 138.67 (Europe, SNP) and 111.15 (Asia, SNP); the above results show that for the outcome, the MR instrumental variable is stronger (Table 3) [40].

Table 3.

Estimated average F-statistic and total variance for each trait

Exposures	Outcome type	Avg F-stat
Body mass index	Malignant neoplasm of colorectal (EUR)	67.64
Body mass index	Malignant neoplasm of colorectal (EAS)	56.17
Body mass index	Serum albumin (EUR)	67.64
Body mass index	Serum albumin (EAS)	56.17
Serum albumin	Malignant neoplasm of colorectal (EUR)	82.87
Serum albumin	Malignant neoplasm of colorectal (EAS)	61.44
Body mass index	Malignant neoplasm of colorectal (EUR)	70.44
Body mass index	Malignant neoplasm of colorectal (EAS)	56.17
Body mass index	C-reactive protein (EUR)	70.40
Body mass index	C-reactive protein (EAS)	56.17
C-reactive protein	Malignant neoplasm of colorectal (EUR)	138.67
C-reactive protein	Malignant neoplasm of colorectal (EAS)	111.15

Regarding the relationship between BMI, Alb, and CRC in the European population, the IVW method showed evidence of the promoting effect of BMI on the risk of CRC, OR: 1.1903, 95% CI 1.0472–1.3530, p < 0.05, β_ALL = 0.1742. The IVW method showed evidence of a positive causal relationship between BMI and Alb, OR: 0.8671, 95% CI 0.8385–0.8967, p < 0.05, β1 = −0.1426. At the same time, the IVW method showed evidence of the protective effect of Alb on CRC risk, OR: 0.8707, 95% CI 0.7590–0.9989, p < 0.05, β2 value=−0.1385. Alb’s direct mediation effect value is 0.154, and the indirect mediation effect value is 0.0198 (Z = 1.92, p = 0.055, St. Error = 0.0103). The proportion of the mediating effect is 11.3% (95%CI 3.23%—22.6%).

The IVW method was used for the East Asian validation population. BMI is associated with an increased risk of CRC, OR: 1.3243, 95% CI 1.0528–1.6658, p < 0.05, β_ALL = 0.2809. A positive causal relationship exists between BMI and Alb, OR: 0.9283, 95% CI: 0.8831–0.9758, p < 0.05, β1 = −0.0744. Alb has a protective effect on the risk of CRC, OR: 0.7157, 95% CI 0.5249–0.9758, p < 0.05, β2 value=−0.3345. The direct mediation effect value of Alb in the East Asian population is 0.2560, and the indirect mediation effect value is 0.0249 (Z = 1.7121, p = 0.0869, Error = 0.0145). The proportion of the mediating effect is 8.86% (95%CI: −1.28%—19.00%). All results are presented in Fig. 2; Table 4.

Fig. 2.

Forest plot presenting the MR estimates for the associations between genetically predicted BMI and CRC risk using European and East Asian ancestry datasets. Results are shown for both univariable MR (IVW random-effects model or IVW) and multivariable MR (MVMR) adjusting for Alb or CRP. Each point represents the estimated OR with corresponding 95% CIs, and the dotted vertical line denotes OR = 1. Detailed results include the number of SNPs used as instruments,β, SE, and P-value for each MR method. Abbreviations: BMI: Body mass index, CRC: Colorectal cancer, Alb: Albumin, CRP: C-reactive protein, MR: Mendelian randomization, IVW: Inverse variance weighted, MVMR: Multivariable Mendelian randomization, SNPs: Single nucleotide polymorphisms,β: Effect estimate, SE: Standard error, OR: Odds ratio,95%LCI: Lower bound of the 95% confidence interval,95%UCI: Upper bound of the 95% confidence interval

Table 4.

Univariable MR analysis of BMI to Abl to CRC

Exposure	Outcome	Exposure ID	Outcome ID	MR method	No SNPs	β	SE	P-value	OR	95%LCI	Steiger P
Body mass index (EUR)	Malignant neoplasm of colorectal (EUR)	GCST90025994	GCST012879	IVW random-effects model	367	0.1742	0.0654	0.0077	1.1903	1.0472	1.16E-82
				MR Egger	367	0.3633	0.1685	0.0317	1.4380	1.0336
				Weighted median	367	0.2018	0.0954	0.0344	1.2236	1.0149
				Simple mode	367	0.2361	0.2676	0.3780	1.2664	0.7495
				Weighted mode	367	0.1536	0.1536	0.3180	1.1660	0.8628
Body mass index (EAS)	Malignant neoplasm of colorectal (EAS)	bbj-a-1	bbj-a-76	IVW random-effects model	65	0.2809	0.1171	0.0164	1.3243	1.0528	7.32E-58
				MR Egger	65	0.4434	0.3422	0.2000	1.5580	0.7966
				Weighted median	65	0.4547	0.1462	0.0019	1.5757	1.1832
				Simple mode	65	0.3760	0.3138	0.2350	1.4564	0.7874
				Weighted mode	65	0.3760	0.2382	0.1900	1.4564	0.9131
Body mass index (EUR)	Serum albumin (EUR)	GCST90025994	GCST90018945	IVW random-effects model	368	−0.1426	0.0171	7.97E-17	0.8671	0.8385	0.00E + 00
				MR Egger	368	−0.1335	0.0444	0.0028	0.8751	0.8022
				Weighted median	368	−0.1341	0.0132	2.00E-24	0.8745	0.8522
				Simple mode	368	−0.1184	0.0352	8.53E-04	0.8883	0.8291
				Weighted mode	368	−0.1325	0.0214	1.62E-09	0.8759	0.8399
Body mass index (EAS)	Serum albumin (EAS)	bbj-a-1	bbj-a-9	IVW random-effects model	65	−0.0744	0.0255	0.0035	0.9283	0.8831	2.44E-202
				MR Egger	65	−0.2282	0.0722	0.0024	0.7960	0.6909
				Weighted median	65	−0.0612	0.0321	0.5660	0.9407	0.8833
				Simple mode	65	−0.0388	0.0657	0.5560	0.9619	0.8457
				Weighted mode	65	−0.0653	0.0573	0.2590	0.9368	0.8372
Serum albumin (EUR)	Malignant neoplasm of colorectal (EUR)	GCST90018945	GCST012879	IVW random-effects model	208	−0.1385	0.0701	0.0482	0.8707	0.7590	1.26E-134
				MR Egger	208	−0.1090	0.1310	0.4060	1.5580	0.7966
				Weighted median	208	−0.1298	0.1024	0.2050	1.5757	1.1832
				Simple mode	208	0.2737	0.2386	0.2530	1.4564	0.7874
				Weighted mode	208	−0.1198	0.1363	0.3810	1.4564	0.9131
Serum albumin (EAS)	Malignant neoplasm of colorectal (EAS)	bbj-a-9	bbj-a-76	Inverse variance weighted	18	−0.3345	0.1582	0.0344	0.7157	0.5249	8.17E-33
				MR Egger	18	−0.0248	0.4577	0.9580	0.9755	0.3978
				Weighted median	18	−0.3531	0.1822	0.0526	0.7025	0.4916
				Simple mode	18	−0.2795	0.2852	0.3410	0.7561	0.4324
				Weighted mode	18	−0.4099	0.2263	0.0877	0.6637	0.4260

1 Cochran Q statistic implemented in MR Egger and IVW method, P > 0·05 indicates no heterogeneity exists. 2. The intercept of MR Egger can be used to indicate whether directional horizontal pleiotropy is driving the results of MR analysis, there are no directional pleiotropies if P > 0·05. 3. MR-PRESSO can detect and adjust for any outliers reflecting horizontal pleiotropic biases,where p value for Global test > 0·05 indicates no horizontal pleiotropic outliers.

At the same time, regarding the relationship between BMI, CRP, and CRC, the IVW method shows that there is evidence that BMI has a weak promoting effect on the risk of CRC in the European population, OR: 1.0013, 95% CI 1.0000–1.0026.0000.0026, p < 0.05, β_ALL = 0.0013. A positive causal relationship exists between BMI and CRP, OR: 1.4306, 95% CI 1.3739–1.4896, p < 0.05, β1 = 0.3581. At the same time, there is evidence that CRP has a weak protective effect on the risk of CRC, OR: 0.9985, 95% CI 0.9971–0.9998, p < 0.05, β2 value=−0.0015. CRP’s direct mediating effect value is −0.0005, and the indirect mediating effect value is 0.0018 (Z=−2.1268, p = 0.0334, St. Error = 0.0003). The proportion of the mediation effect of CRP is −41.32% (95%CI −79.40% to −3.24%).

In the East Asian validation population, BMI is associated with an increased risk of CRC, OR: 1.3243, 95% CI 1.0528–1.6658, p < 0.05, β_ALL = 0.2809. There is a positive causal relationship between BMI and CRP, OR: 1.1008, 95% CI 1.0496–1.1544, p < 0.05, β1 = 0.0960, but there is no evidence of the causal relationship between CRP and CRC risk, OR: 0.9992, 95% CI 0.7628–1.3087, p > 0.05, β2 value=−0.0008. CRP’s direct mediation effect value in the East Asian population is −7.68e-05, and the indirect mediation effect value is 0.2810 (Z=−0.00581, p = 0.9954, St. Error = 0.0132). There is no mediating effect. All results are presented in Fig. 2; Table 5.

Table 5.

Univariable MR analysis of BMI to CRP to CRC

Exposure	Outcome	Exposure ID	Outcome ID	MR method	No SNPs	β	SE	P-value	OR	95%LCI	Steiger P
Body mass index (EUR)	Malignant neoplasm of colorectal (EUR)	ieu-b-40	ukb-d-C18	Inverse variance weighted	495	0.0013	0.0007	0.0462	1.0013	1.0000	0.00E + 00
				MR Egger	495	0.0038	0.0019	0.0407	1.5580	0.7966
				Weighted median	495	0.0012	0.0010	0.2500	1.5757	1.1832
				Simple mode	495	0.0008	0.0035	0.8150	1.4564	0.7874
				Weighted mode	495	0.0001	0.0028	0.9830	1.4564	0.9131
Body mass index (EAS)	Malignant neoplasm of colorectal (EAS)	bbj-a-1	bbj-a-76	IVW random-effects model	65	0.2809	0.1171	0.0164	1.3243	1.0528	7.32E-58
				MR Egger	65	0.4434	0.3422	0.2000	1.5580	0.7966
				Weighted median	65	0.4547	0.1462	0.0019	1.5757	1.1832
				Simple mode	65	0.3760	0.3138	0.2350	1.4564	0.7874
				Weighted mode	65	0.3760	0.2382	0.1190	1.4564	0.9131
Body mass index (EUR)	C-reactive protein (EUR)	ieu-b-40	ieu-b-35	IVW random-effects model	494	0.3581	0.0206	1.89E-67	1.4306	1.3739	0.00E + 00
				MR Egger	494	0.4017	0.0590	2.97E-11	1.4944	1.3311
				Weighted median	494	0.3647	0.0201	1.30E-73	1.4401	1.3844
				Simple mode	494	0.2975	0.0680	1.48E-05	1.3465	1.1785
				Weighted mode	494	0.4070	0.0421	2.13E-20	1.5023	1.3835
Body mass index (EAS)	C-reactive protein (EAS)	bbj-a-1	bbj-a-14	Inverse variance weighted	65	0.0960	0.0243	7.68E-05	1.1008	1.0496	8.93E-168
				MR Egger	65	−0.0198	0.0702	0.7790	0.9804	0.8544
				Weighted median	65	0.0812	0.0386	0.0356	1.0846	1.0055
				Simple mode	65	0.1083	0.0890	0.2280	1.1144	0.9361
				Weighted mode	65	0.0787	0.0573	0.1740	1.0819	0.9670
C-reactive protein (EUR)	Malignant neoplasm of colorectal (EUR)	ieu-b-35	ukb-d-C18	Inverse variance weighted	53	−0.0015	0.0007	0.0231	0.9985	0.9971	0.00E + 00
				MR Egger	53	−0.0013	0.0011	0.2500	0.9987	0.9965
				Weighted median	53	−0.0011	0.0009	0.2470	0.9989	0.9971
				Simple mode	53	0.0006	0.0018	0.7600	1.0006	0.9970
				Weighted mode	53	−0.0012	0.0008	0.1560	0.9988	0.9972
C-reactive protein (EAS)	Malignant neoplasm of colorectal (EAS)	bbj-a-14	bbj-a-76	Inverse variance weighted	7	−0.0008	0.1377	0.9950	0.9992	0.7628	2.72E-42
				MR Egger	7	0.1250	0.3285	0.7190	1.1331	0.5952
				Weighted median	7	−0.0362	0.1770	0.8380	0.9644	0.6817
				Simple mode	7	0.1737	0.3046	0.5890	1.1897	0.6549
				Weighted mode	7	−0.1582	0.2268	0.5110	0.8536	0.5473

1 Cochran Q statistic implemented in MR Egger and IVW method,P > 0·05 indicates no heterogeneity exists. 2. The intercept of MR Egger can be used to indicate whether directional horizontal pleiotropy is driving the results of MR analysis,there are no directional pleiotropies if P > 0·05. 3. MR-PRESSO can detect and adjust for any outliers reflecting horizontal pleiotropic biases,where p value for Global test > 0·05 indicates no horizontal pleiotropic outliers.

Table S1 shows that there is no forward or reverse causal relationship between CRP and Alb in both the European and East Asian populations. The instrumental variables used for each exposure–outcome pair are presented in Table S2.

Sensitivity analyses

After calculating the MR results from BMI to Abl and CRC, the next step is to perform a sensitivity analysis. First, a heterogeneity test: Regardless of whether it is European or East Asian populations, the MR results from BMI to Abl and CRC show obvious heterogeneity (Het p < 0.05). The IVW random-effects model was used in the final statistical model. Pleiotropy test: The MR-PRESSO method identifies SNP-level pleiotropy outliers (p < 0.05). After these outliers are removed, the intercept term of the MR Egger method indicates that the intercept is slightly different from 0 (p > 0.05), suggesting no horizontal pleiotropy. This implies that the instrumental variables do not significantly affect the results in other ways, except through exposure. Sensitivity analysis of BMI to CRP and CRC. The Asian population’s BMI to CRC and the European population’s BMI to CRP showed significant heterogeneity (Heterogeneity p < 0.05). The IVW random-effects model was used in the final statistical model. Others did not show heterogeneity, so the Inverse variance-weighted statistical model was selected. In the pleiotropy test, the MR-PRESSO method identified the existence of SNP-level pleiotropy outliers (p < 0.05). After these outliers were removed, the intercept term of the MR Egger method was slightly different from 0 (p > 0.05), indicating no horizontal pleiotropy.

Finally, Steiger filtering analysis provided strong evidence that the causal direction between BMI, CRP, Alb genetic susceptibility, and CRC is from the former to the latter (Steiger Correct direction is TRUE, Steiger p < 0.05). All results are available in Table 6.

Table 6.

Sensitivity analysis

Causal path	Exposure	Outcome	Exposure ID	Outcome ID	MR method	Heterogeneity test			Pleiotropy test		MR-PRESSO	Steiger correct direction	Steiger P
						Cochran’s Q	Het P	No SNPs removed	egger intercept	Ple P	P
BMI to Abl to CRC	Body mass index (EUR)	Malignant neoplasm of colorectal (EUR)	GCST90025994	GCST012879	IVW random-effects model	491.2263	1.28E-05	1	−0.0037	0.2243	< 0.001	TRUE	1.16E-82
					MR Egger	489.2403	1.43E-05	2
	Body mass index (EAS)	Malignant neoplasm of colorectal (EAS)	bbj-a-1	bbj-a-76	IVW random-effects model	101.6623	1.91E-03	1	−0.0053	0.6147	0.0030	TRUE	7.32E-58
					MR Egger	101.2511	1.60E-03	2
	Body mass index (EUR)	Serum albumin (EUR)	GCST90025994	GCST90018945	IVW random-effects model	2162.4236	1.79E-251	1	−0.0002	0.8239	< 0.001	TRUE	0.00E + 00
					MR Egger	2162.1305	8.31E-252	2
	Body mass index (EAS)	Serum albumin (EAS)	bbj-a-1	bbj-a-9	IVW random-effects model	97.9972	4.01E-03	1	0.0050	0.0270	0.0030	TRUE	2.44E-202
					MR Egger	90.6185	0.0129	2
	Serum albumin (EUR)	Malignant neoplasm of colorectal (EUR)	GCST90018945	GCST012879	IVW random-effects model	289.8955	1.25E-04	1	−0.0009	0.7905	0.0020	TRUE	1.26E-134
					MR Egger	289.7960	1.06E-04	2
	Serum albumin (EAS)	Malignant neoplasm of colorectal (EAS)	bbj-a-9	bbj-a-76	Inverse variance weighted	23.1529	0.1440	1	−0.0146	0.4804	0.1850	TRUE	8.17E-33
					MR Egger	22.4215	0.1300	2
BMI to CRP to CRC	Body mass index (EUR)	Malignant neoplasm of colorectal (EUR)	ieu-b-40	ukb-d-C18	Inverse variance weighted	544.6334	0.0571	1	0.0000	0.1491	0.0540	TRUE	0.00E + 00
					MR Egger	542.3368	0.0615	2
	Body mass index (EAS)	Malignant neoplasm of colorectal (EAS)	bbj-a-1	bbj-a-76	IVW random-effects model	101.6623	1.91E-03	1	−0.0053	0.6147	0.0030	TRUE	7.32E-58
					MR Egger	101.2511	1.60E-03	2
	Body mass index (EUR)	C-reactive protein (EUR)	ieu-b-40	ieu-b-35	IVW random-effects model	1541.7210	2.54E-108	1	−0.0007	0.4307	< 0.001	TRUE	0.00E + 00
					MR Egger	1539.7744	2.79E-108	2
	Body mass index (EAS)	C-reactive protein (EAS)	bbj-a-1	bbj-a-14	Inverse variance weighted	65.6261	0.4200	1	0.0038	0.0841	0.4560	TRUE	8.93E-168
					MR Egger	62.5449	0.4920	2
	C-reactive protein (EUR)	Malignant neoplasm of colorectal (EUR)	ieu-b-35	ukb-d-C18	Inverse variance weighted	65.7688	0.0950	1	0.0000	0.7642	0.0950	TRUE	0.00E + 00
					MR Egger	65.6517	0.0814	2
	C-reactive protein (EAS)	Malignant neoplasm of colorectal (EAS)	bbj-a-14	bbj-a-76	Inverse variance weighted	5.5222	0.4790	1	−0.0095	0.6885	0.4760	TRUE	2.72E-42
					MR Egger	5.3297	0.3770	2

Multivariable MR analysis of BMI, Alb, and CRP on CRC

MVMR was used to perform mediation analysis to determine any putative genetic effects on CRC risk. All results are available in Table 6. We found that the association between BMI and CRC risk in the European population, via a direct effect (unadjusted OR 1.1903 [1.0472–1.3530], p = 0.0077), was slightly altered after adjustment for Alb (adjusted OR 1.1639 [1.0082–1.3437], p = 0.0383). The proportion is 12.6%. The association between BMI and CRC risk in the Asian population, through direct effects (unadjusted OR 1.3243 [1.0528–1.6658], p = 0.0164), slightly changed after adjustment for Alb (adjusted OR 1.3390 [1.0595–1.6921], p = 0.0145); the proportion is 8%. Secondly, CRP was used to explore the potential mediating effect. Because the two-step MR analysis showed that CRP was directly positively related to BMI and directly negatively related to CRC risk, the association between BMI and CRC risk in the European population was through the direct effect (unadjusted OR 1.0013[1.0000–1.0026.0000.0026], p = 0.0462) changed slightly after adjusting for CRP (adjusted OR 1.0016 [1.0002–1.0030], p = 0.0383), proportion is −36.5%. The association between BMI and CRC risk in the Asian population, through direct effects (unadjusted OR 1.3243 [1.0528–1.6658], p = 0.0164), was slightly altered after adjustment for CRP (adjusted OR 1.4030 [1.1143–1.7664], p = 0.0145), with a proportionate change of −0.9%. This suggests a possible partial mediating role for the two risk factors considered. All results are available in Tables 7 and 8, and Table S3. The results calculated by univariable and multivariable analyses are consistent.

Table 7.

Results of mediator variable for BMI and CRC

		Univariable analysis											MVMR analysis
		βALL	β1	SE1	β2	SE2	Mediating effect	Direct effect	Sobel test Z	Sobel test p	St. Error	proportion	β3	β4	Mediating effect	Direct effect	proportion
BMI as the exposure, CRC as the outcome, and Alb as the mediator	European	0.1742	−0.1426	0.0171	−0.1385	0.0701	0.0198	0.1540	1.9225	0.0545	0.0103	11.3%(3.23%−22.6%)	−0.1534	0.1518	0.0219	0.1518	12.6%
	East Asia	0.2809	−0.0744	0.0255	−0.3345	0.1582	0.0249	0.2560	1.7121	0.0869	0.0145	8.86%(−1.28%−19.00%)	−0.3404	0.2919	0.0253	0.2919	8.0%
BMI as the exposure, CRC as the outcome, and CRP as the mediator	European	0.0013	0.3581	0.0206	−0.0015	0.0007	−0.0005	0.0018	−2.1268	0.0334	0.0003	−41.32%(−79.40%-−3.24%)	−0.0012	0.0016	−0.0004	0.0016	−36.5%
	East Asia	0.2809	0.0960	0.0243	−0.0008	0.1377	−0.0001	0.2810	−0.0058	0.9954	0.0132	−0.027%(−9.251%-−9.196)	−0.0309	0.3386	−0.0030	0.3386	−0.9%

Table 8.

MVMR analysis of BMI, Alb, CRP, and CRC

Exposure	Outcome	Method	No SNPs	β	SE	P-value	pval_adjusted*	OR	95%LCI	95%UCI
Albumin|| id: ebi-a-GCST90018945	Colorectal cancer || id: ebi-a-GCST012879	IVW MVMR (direct)	88	−0.1534	0.0874	0.0793	0.0793	0.8578	0.7227	1.0181
Body mass index || id: ebi-a-GCST90025994		IVW MVMR (direct)	250	0.1518	0.0733	0.0383	0.0766	1.1639	1.0082	1.3437
Albumin || id: bbj-a-9	Colorectal Cancer || id: bbj-a-76	IVW MVMR (direct)	12	−0.3404	0.1735	0.0498	0.0498	0.7115	0.5064	0.9997
Body mass index || id: bbj-a-1		IVW MVMR (direct)	51	0.2919	0.1194	0.0145	0.0290	1.3390	1.0595	1.6921
C-Reactive protein level || id: ieu-b-35	Diagnoses - main ICD10: C18 Malignant neoplasm of colon || id: ukb-d-C18	IVW MVMR (direct)	25	−0.0012	0.0006	0.0390	0.0390	0.9988	0.9977	0.9999
body mass index || id: ieu-b-40		IVW MVMR (direct)	451	0.0016	0.0007	0.0229	0.0390	1.0016	1.0002	1.0030
C-reactive protein || id: bbj-a-14	Colorectal Cancer || id: bbj-a-76	IVW MVMR (direct)	4	−0.0309	0.1835	0.8663	0.8663	0.9696	0.6766	1.3894
Body mass index || id: bbj-a-1		IVW MVMR (direct)	51	0.3386	0.1175	0.0040	0.0079	1.4030	1.1143	1.7664

*Benjamini-Hochberg adjusted P-value threshold

Cohort observational analysis between BMI, CRP, Alb, and CRC

After 477 passed the screening and selection criteria, 374 local patients remained (Table 1). A total of 85 patients ultimately experienced the time-to-event outcome, namely death from any cause. Univariate Cox regression analysis found that BMI, Alb, and CRP (HR: 0.89, 0.92, and 1.02, respectively) were all associated with shorter OS of CRC (Table S4). In the minimally adjusted model “Model 1”-BMI-adjusted Alb, BMI was inversely associated with poor CRC prognosis (HR: 0.91, 95% CI 0.85–0.98, P < 0.05), “Model 2” - BMI adjusts CRP, BMI is not related to shorter OS of CRC (P > 0.05), “Model 3” - Alb adjusts CRP, Alb is not related to shorter OS of CRC (P > 0.05), “Model 4” - BMI, Alb, and CRP Only CRP was positively associated with shorter OS in CRC (HR: 1.02, 95% CI 1.02–1.03, P < 0.05).“Model 5” - After combining age and tumor marker proteins CEA and CA199, CRP is positively correlated with shorter OS of CRC (HR: 1.02, 95% CI 1.01–1.03, P < 0.05), and BMI is negatively correlated with shorter OS of CRC. (HR: 0.92, 95% CI 0.85–0.99, P < 0.05).“Model 6” - After combining age, tumor marker proteins, and peripheral blood leukocytes, CRP is positively associated with shorter OS of CRC (HR: 1.02, 95% CI 1.01–1.03, P < 0.05), and BMI is negatively associated with shorter OS of CRC. Correlated (HR: 0.91, 95% CI 0.84–0.99, P < 0.05); detailed data are shown in Fig. 3.

Fig. 3.

Retrospective analysis estimates the effects of BMI after adjust Alb, CRP and other variables on CRC Prognosis outcomes. Forest plot showing hazard ratios (HRs) and 95% confidence intervals (CIs) for the associations between clinical characteristics and overall survival across six multivariable Cox regression models. Each model includes different combinations of covariates, and statistically significant estimates are highlighted. The dotted vertical line represents HR = 1.0. Each Model Annotation: Model 1(BMI-adjusted Alb model), Model 2(BMI-adjusted CRP model), Model 3(Alb-adjusted CRP model), Model 4(BMI + Alb + CRP model), Model 5(BMI + Alb + CRP + age + tumor markers model), Model 6(BMI + Alb + CRP + age + tumor markers + peripheral leukocytes model).Abbreviations: BMI: Body mass index, Alb: Albumin, CRP: C-reactive protein, WBC: White blood cell count, NEU%: Neutrophil percentage, NEU#: Neutrophil count, MONO#: Monocyte count, LYM%: Lymphocyte percentage, BASO#: Basophil count, LMR: Lymphocyte-to-monocyte ratio, CEA: Carcinoembryonic antigen, CA199: Carbohydrate antigen 19 − 9, HR: Hazard ratio, CI: Confidence interval

Discussions

This study, along with most others, concludes that high BMI is significantly associated with an increased incidence of colorectal cancer [3, 41]. Chronic low-grade inflammation resulting from high BMI may promote cancer development through various mechanisms [42]. This includes producing large amounts of pro-inflammatory cytokines, such as IL-6 and TNF-α. These cytokines can induce and sustain chronic inflammation in the intestinal microenvironment, thereby increasing the risk of colorectal cancer [43]. Additionally, it may promote the occurrence and development of cancer by affecting insulin resistance, insulin-like growth factor (IGF) levels, and hormone metabolism [42]. Our findings are consistent with the growing evidence that systemic inflammatory and nutritional biomarkers play a crucial role in predicting survival in colorectal cancer. Moreover, several inflammation- and nutrition-based prognostic scoring systems have been developed or validated for stage I–III CRC, reinforcing the importance of biomarkers such as CRP and Alb in risk stratification [19, 21]. In addition, systemic biochemical markers, including serum calcium levels, have also been associated with CRC outcomes [22], further supporting the prognostic relevance of host nutritional and inflammatory status. Taken together, these studies support our observations that CRP and Alb contribute meaningfully to CRC prognosis, underscoring their potential value in clinical assessment and the development of individualized management strategies.

Serum Alb and CRP are important inflammatory markers [44]. In a state of chronic inflammation, the protein synthesis capacity of the liver is inhibited, and serum Alb levels are reduced [45]. CRP is an acute-phase response protein synthesized by the liver [46]. Sustained elevated CRP levels may indicate the presence of chronic inflammation, such as inflammatory bowel disease [47]. Low serum Alb and high CRP levels often indicate chronic inflammation and underlying malnutrition. In some diseases (such as colorectal cancer, chronic liver disease, etc.), the combined detection of ALB and CRP can be used to judge prognosis[48, 49]. Low ALB and high CRP levels are generally associated with a poor prognosis[50, 51] . In our study, the value of inflammatory markers in predicting the survival and prognosis of CRC was also evaluated. and that chronic inflammation itself is a cause and key mediator of key factors in the development of CRC. Additionally, multiple studies have used MR analysis to confirm a causal relationship between high BMI, particularly in early childhood and adulthood, and an increased risk of CRC [52–54]. Some studies have differentiated the effects of overall and abdominal fat accumulation, finding that abdominal fat accumulation has a more significant impact on CRC [55]. At the same time, TMR studies based on basal metabolic rate have demonstrated a causal relationship between a high metabolic rate and an increased risk of CRC [56]. Childhood fat accumulation affects overall health in adulthood and may increase the incidence of CRC later in life [52]. In our study, the positive causal relationship between BMI and an increased risk of colorectal cancer was confirmed in both European and East Asian populations. Studies have identified an association between serum Alb levels and colorectal cancer risk, though the exact cause-and-effect relationship remains unclear [57]. Some studies have shown that lower Alb levels are associated with an increased risk of colorectal cancer, but the evidence for causality remains inconsistent [50]. In our study, Alb levels were inversely associated with CRC risk in both European and East Asian populations, consistent with observational studies on prognosis. Multiple studies have used Mendelian Randomization (MR) analysis to confirm the causal relationship between elevated CRP levels in the blood and an increased risk of CRC. This supports the significant role of inflammation in cancer development, suggesting that controlling inflammation levels may reduce the incidence of CRC [9, 51, 58]. Our research also confirmed this. Additionally, multiple studies have employed MR analysis to demonstrate a causal relationship between increased BMI and elevated CRP levels [9, 25]. Our study found that BMI and CRP levels exhibited a weak negative correlation and were not significant in East Asian populations. It may be necessary to verify this in multiple datasets later. While no direct research literature delineates the causal relationship between Alb and BMI, our findings suggest that Alb levels are closely associated with overall health and nutritional status, and that BMI exhibits an inverse relationship with Alb levels.

There is a complex relationship between BMI and CRC, influenced by numerous intermediary factors, including CRP, adiponectin, leptin, resistin, high-density lipoprotein cholesterol (HDL-C), and low-density lipoprotein cholesterol (LDL-C) [59]. CRP mediates various diseases, primarily through the inflammatory pathway. Specifically, CRP mediates the relationship between BMI and conditions such as cardiovascular disease, type 2 diabetes, cancer, chronic kidney disease, chronic obstructive pulmonary disease, osteoarthritis, and rheumatoid arthritis [60–63]. Chronic inflammation caused by a high BMI increases CRP levels, and elevated CRP levels exacerbate the inflammatory response in these diseases [62]. Consistent with our study. However, the relationship between CRP and CRC risk remains controversial, and our MR study of East Asian populations did not find any correlation. Alb has antioxidant effects, neutralizing free radicals in the tumor microenvironment and inhibiting tumor growth [64]. Low Alb levels may impair this protective effect. It also plays a key role in mediating the relationship between BMI and various diseases [65–68]. One explanation is that high BMI leads to fatty liver disease and decreased liver function, which in turn leads to lower Alb levels [69]. Additionally, a high BMI is linked to chronic low-grade inflammation, and the rise in inflammatory factors, such as CRP, may directly impact Albumin levels [70]. Low Alb levels, in turn, aggravate the inflammatory response, forming a vicious cycle and further promoting the occurrence and development of cancer. However, our study found no direct evidence of a causal relationship between CRP and Alb. CRP and Alb have independent mediating effects on BMI in CRC. This was also confirmed in our retrospective cohort of CRC. In our CRC retrospective study cohort, BMI, Alb, and CRP contributed to CRC prognosis in univariate analysis. Still, BMI and Alb had no statistically significant impact on CRC prognosis in a multifactor analysis adjusting for CRP; it is speculated that BMI, Alb, and CRC prognosis may be false associations or indirect influencing factors. Therefore, after adjusting for the impact of CRP, this “false association” between BMI and CRC disappeared. The genetic results are consistent with the finding that CRP is a mediating factor between BMI and CRC. The contributions of BMI and Alb to the prognosis of CRC are independent of each other. Low Alb levels and inappropriate BMI (too high or too low) may lead to shorter OS. BMI statistically affects prognosis after adjusting for CRP, age, tumor protein markers, and peripheral blood leukocytes. Age, tumor protein markers, and peripheral blood may strengthen the association between BMI and CRC prognosis. In observational cohort analyses, the temporal order between exposure and outcome cannot be fully established and may be influenced by the underlying disease process. As demonstrated in our study, elevated CRP levels may not be the cause of CRC but rather a consequence of early tumor–related inflammatory responses. Similarly, decreased albumin levels may result from cancer-induced malnutrition or chronic inflammation, rather than low albumin being a causal factor for cancer. Therefore, in our observational cohort analysis, reverse causation may arise because specific exposures (such as CRP or albumin) can be affected by the preclinical or early stages of the disease, leading to potential misinterpretation of the temporal relationship between exposure and outcome. Additionally, it appears that the influence of peripheral blood leukocytes on CRC prognosis could be related to BMI.

Limitations

First, the actual causal path may be more complex than our model assumes, and there may be multiple mediating factors or interactions, making it difficult to accurately identify and quantify the specific role of each mediating factor. In some cases, the reverse causal relationship between BMI, CRC, CRP, and Alb (i.e., CRC affects CRP) may be confused with the forward causal relationship (where the mediating factor, CRP, affects CRC). Second, although mediation analysis can control for some confounders, unadjusted confounders may still exist, particularly in observational studies, which can potentially impact the accuracy of the mediation effect. Third, our study employed a single mediator model; however, multiple mediator variables and complex interactions may exist, and a single model may not be sufficient to capture these complexities. We utilized European and East Asian populations to maximize protection against situations where results in a specific population or setting may not necessarily generalize to other populations or settings, but this is not comprehensive coverage. In our study, the phenotype ‘diagnoses – main ICD10: C18’ (ukb-d-C18) corresponds only to malignant neoplasm of the colon, which also introduces certain limitations. Finally, it is essential to note that some odds ratios in our MR analyses were very close to unity. Although statistically significant due to the large sample size, such small effect sizes should be interpreted cautiously, as they may indicate limited biological or clinical relevance.

Conclusions

The significance of mediating effects is crucial in studying the relationship between BMI and colorectal cancer, as it helps to uncover how BMI influences disease risk through specific biological or behavioral pathways. For example, by understanding the mediating roles of Alb and C-reactive protein in the relationship between BMI and CRC, we can determine how high BMI increases cancer risk through mechanisms such as inflammation and immune function. Once mediating factors are identified, interventions can be targeted at these factors. For example, the impact of BMI on CRC can be altered by improving nutritional status (increasing Alb levels) and controlling inflammation (lowering CRP levels). This targeted intervention is more effective than weight management alone. Considering intermediary factors such as Alb and CRP, a comprehensive treatment plan can be formulated that not only focuses on the cancer itself but also considers the patient’s nutritional and inflammatory status, thereby improving treatment effects and prognosis.

Author contributions

Study design: HL, MX. Data acquisition: HL. Quality control of data and algorithms: HL. Data analysis and interpretation: XX, ZC, JS. Statistical analysis: HL, XX. Manuscript preparation: XX, ZC, JS, CH, CS, HL, QT; Manuscript editing: HL, MX. All authors reviewed and approved the final manuscript.

Funding

This study was supported by grants from the Project supported by the Joint Funds of the Natural Science Foundation of Hunan Province (Grant No.2023JJ50381); the Self-funded scientific research project of the Guangxi Zhuang Autonomous Region Health Committee (Grant No. Z-A20230756,Z-A20240733).

Data availability

This published article and its supplementary information files include all data generated or analyzed during this study. The corresponding author can provide all possible assistance to the requester of the original data.

Declarations

Ethics approval and consent to participate

The Guangxi Medical University Cancer Hospital Ethics Committee reviewed and approved the studies involving human participants. The patients/participants provided written informed consent to participate in this study.

Consent for publication

All authors agree to the publication of this article.

Competing interests

The authors declare no competing interests.