Diagnostic performance of serum metabolites biomarker associated with colorectal adenoma: a systematic review

Search strategy

Literature searches were performed in three electronic databases: PubMed, Web of Science (WoS), and Scopus to obtain a comprehensive finding of unique citations of related articles. Three databases were selected to increase the validity and quality of the results and to minimize selection bias as suggested by Vassar et al. (2017). The search strategy was designed by three reviewers (MFAB, AMN, and SM) and conducted by MFAB, AMN, CSF, TJK, and SM. Searching keywords including: “metabolites” OR “metabolism product” OR “metabolic product” OR “metabolomics” or “metabolome” AND “marker” OR “biomarker” OR “biological marker” OR “biological signature” AND “blood” OR “serum” OR “circulating” AND “human” AND “Sensitivity” AND “Specificity” AND “diagnosis” OR “screening” OR “testing” OR “detecting” AND “Colorectal adenoma” OR “Colonic Polyps” OR “Polyps”. Two reviewers (MFAB and AMN) independently assessed titles and abstracts of all abstracts as part of the primary screen. A secondary screen of titles and abstracts was then conducted by a further two reviewers (SM and CSF). The results were analyzed by MFAB, SM, and AMN and confirmed by TJK and CSF.

Eligibility criteria

Studies published between 2014 and 2023, were included to ensure that all newly published evidence on diagnostics accuracy of potential serum biomarkers for colorectal adenoma screening. The review was limited to studies on serum samples from humans, published in English, and addressed the finding of serum metabolite biomarkers in detecting CRA with sensitivity and specificity.

Exclusion criteria

Review articles, conference abstracts, studies without a complete set of data, and articles that did not mention CRA, or serum metabolites in the title or abstract were excluded. In addition, the study was limited to serum biomarkers as this type of sample is easily obtained, hence, all other sources of CRA metabolites biomarkers such as stool or urine, were excluded. Studies with no definition of the role of serum biomarkers in colorectal adenoma diagnosis were also excluded.

Quality assessment

The quality of each publication was evaluated by two independent reviewers (MFAB and AMN) and confirmed by the other two authors (SZ and CSF). QUADOMICS was used to assess the methodological quality of the selected studies. The quality of the studies was summarized by the percentage of applied criteria scored positively (Table S1).

Data extraction

All articles were screened by two authors (MFAB, and AMN) and any disagreement was reached by consensus or involvement of CSF and SM. Data were extracted by two authors (MFAB, and AMN) followed by validation by SM, TJK, and CSF. The articles/studies were selected based on inclusion and exclusion criteria.

The following information was extracted from all included studies: First named author, year of publication, participant country, sample type used, number of participants, suggested serum metabolites associated with CRA, sensitivity, specificity, and type of analysis instrument as suggested by Contreras et al. (2023).

Data synthesis

Identified serum metabolite biomarkers for CRA were extracted from included studies. Enrichment pathway analysis was conducted using MetaboAnalyst 5.0 for all individual/panels of metabolites reported for CRA. The sensitivity and specificity of the serum biomarkers from the included studies were extracted and logistically transformed, and then a linear regression line was fitted through the data points. This line was back-transformed to obtain the ROC curve, which is a compact description of the accuracy of the diagnostic test in many populations.

Study selection

A PRISMA diagram of studies selected for this systematic review is summarized in Fig. 1. The search strategy identified 4,212 suitable abstracts, from which 4,198 were excluded by review of the title and abstract during the primary and secondary screens, as they did not meet the eligibility criteria. Full-text articles were obtained for 14 studies. A total of nine articles (Guo et al., 2023; Tevini et al., 2022; Gu et al., 2019; Liu et al., 2023; Chen et al., 2017; Long et al., 2017; Uchiyama et al., 2017; Farshidfar et al., 2016; Zhu et al., 2014), examining potential serum metabolites biomarker for CRA were included in this review for data extraction and analysis (Table 1).

Three of the studies (Tevini et al., 2022; Liu et al., 2023; Chen et al., 2017) used the cohort study design while the remaining six studies (Guo et al., 2023; Gu et al., 2019; Long et al., 2017; Uchiyama et al., 2017; Farshidfar et al., 2016) used a case-control. These studies were conducted in the USA, China, Japan, Canada, and Austria with a range of 8-320 participants for each group using six different analytical platforms, including LC-MS, Isoberic Labelling-MS, UHPLC-MS/MS, GC-MS, 1H-NMR, and FIA and LC-MS (Table 1). Only three studies (Gu et al., 2019; Uchiyama et al., 2017; Zhu et al., 2014) reported sensitivity and specificity of suggested serum biomarkers (Table 1).

The quality of studies

The quality assessment results for the individual studies were conducted using QUADOMICS (Table S1). All studies included in this review met the inclusion criteria and scored positively which indicates that the overall quality was good.

Serum metabolites biomarkers differentiating colorectal adenomas vs. normal

Seven studies (Guo et al., 2023; Tevini et al., 2022; Liu et al., 2023; Chen et al., 2017; Uchiyama et al., 2017; Farshidfar et al., 2016; Zhu et al., 2014) had reported a list of potential individual/panel serum metabolites for CRA vs. Normal. However, only two of them have reported the sensitivity/specificity value (Table 2). The differential serum metabolites associated with CRA vs. normal identified by seven included studies in Table 1 were enriched into pathways analysis. The five most affected pathways in CRA vs. normal as presented in Fig. 2 were the glycoxylate and dicarboxylate metabolism; alanine, aspartate, and glutamate metabolism; aminoacyl-tRNA biosynthesis; D-glutamine and D-glutamate metabolism; and nitrogen metabolism.

The individual sensitivity and specificity values for each serum metabolite were further extracted (Table 3) and analyzed by constructing a graph of sensitivity vs. false positive rates (FPR) to identify their diagnostic accuracy. Glycerol and citrate were excluded from the list due to the unavailability of their individual data. Figure 3 shows the graph of sensitivity vs. FPR for these biomarkers where all three serum metabolites (benzoic acid, acetate, and lactate) show significant diagnostic accuracy in differentiating CRA and normal patients (AUC > 0.7).

Serum metabolites biomarker differentiating CRA vs. CRC

Seven studies (Guo et al., 2023; Tevini et al., 2022; Gu et al., 2019; Chen et al., 2017; Long et al., 2017; Uchiyama et al., 2017; Farshidfar et al., 2016) reported individual/panel serum biomarkers associated with CRA vs. CRC. The differential serum metabolites reported by these studies (Table 1) were enriched into pathways analysis. Figure 4 presents the affected pathways for CRA vs. CRC. The five most affected pathways were pyruvate metabolism, glycolysis/gluconeogenesis, glycerolipid metabolism, citrate/TCA cycle, and alanine, aspartate, and glutamate metabolism.

Only three of the studies (Gu et al., 2019; Uchiyama et al., 2017; Zhu et al., 2014) have included the value of sensitivity and specificity for each reported serum biomarker CRA vs. CRC (Table 4). The individual sensitivity and specificity values for each serum metabolite were further extracted (Table 5) and analyzed by constructing a graph of sensitivity vs. FPR to identify their diagnostic accuracy. Three of them (adenosine, pentothenate, and linoleic acid) show AUC >0.9 (Fig. 5).

Limitations

A small number of reviewed studies reporting the sensitivity and specificity of serum metabolites associated with CRA indicates a huge gap in studying serum metabolite biomarkers for CRA detection. Different analytical platforms and diverse populations in reviewed studies led to differences in signature serum metabolites identified between studies. Although a limited number of studies provided a limited data analysis of diagnostic accuracy for these serum metabolites, this review could serve as a reference for future research in selecting the best candidates of serum metabolites with reliable accuracy in detecting CRA.