Introduction
Colorectal cancer (CRC) is the third most common malignancy worldwide. According to global statistics, 1.93 million new cases were diagnosed in 2020 [
1]. Despite advances in treatment, including surgical skills, chemotherapy regimens, and the use of biological agents, the rate of death due to CRC remains higher than that due to other cancers, with an estimated 930,000 deaths and CRC was the second-leading cause of cancer-related deaths in 2020 [
1]. Because of the heterogeneous nature of CRC, the treatment strategy and outcomes are diverse, even for tumors with the same stage [
2]. A 2018 study that assessed the global patterns and trends in CRC demonstrated a diversity in incidence and a mortality rate up to 10-fold worldwide, with distinct gradients across different levels of human development [
3]. Determining the prognostic factors for CRC is a critical issue for making treatment plans for CRC and establishing guidance on how intensely the patient should be followed after index treatment.
It is well known that cancer-related inflammation can aid tumor proliferation, angiogenesis, progression, metastasis, and resistance to chemotherapy [
4,
5]. Based on this concept, several serum inflammatory markers have been reported as prognostic biomarkers in different cancer types [
6]. Either the solitary or combined use of neutrophils, lymphocytes, monocytes, platelets, C-reactive protein (CRP), and albumin has been associated with CRC survival [
7‐
11]. These markers can be further classified into cellular and protein components. CRP is an acute-phase protein regulated by several pro-inflammatory cytokines such as tumor necrosis factor-alpha, interleukin-1, and interleukin-6 [
12]. Albumin is recognized as a nutritional status parameter and is associated with chronic inflammation [
13,
14]. The utility of these two protein markers has been classically reported as the Glasgow prognostic score (GPS) and modified Glasgow prognostic score (mGPS). GPS and mGPS have been reported as independent prognostic factors for CRC [
15,
16]. Recently, a novel inflammation-based marker composed of the CRP-to-albumin ratio (CAR) was proposed to predict the outcome of patients with severe sepsis [
17] and served as a prognostic factor for a variety of human malignancies [
18]. However, there are conflicting results regarding the prognostic value of CAR in CRC. Ishizuka et al. noted that the overall survival had significantly increased in the low pre-treatment CAR subgroup compared to that of the high pre-treatment CAR subgroup in their study of 627 patients with colorectal cancer who had undergone elective surgery (HR 2.596,
P < 0.001) [
19]. On the contrary, Zhou et al. demonstrated that the dynamic change in CAR pre- and post-surgery was strongly associated with the overall survival in patients with CRC who had undergone surgery; however, if only the pre-operative CAR data was used, it did not significantly affect the overall survival (HR 1.59, 95% CIs 0.86–2.91) [
20]. Previous studies analyzing the correlation between pre-treatment CAR and survival in CRC mainly came from East Asia; therefore, evidence about the application of this biomarker in the Western countries is lacking. Thus, we conducted this updated meta-analysis to assess the association between the pre-treatment value of CAR and outcomes in CRC patients by reviewing the findings of all recently published studies. Moreover, we aimed to identify the clinicopathological features associated with a high CAR.
Methods
Search strategy
This study was conducted according to the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-analysis [
21]. A comprehensive search was conducted using PubMed and Embase databases from the earliest records to October 2020. The search terms were as follows: (C-reactive protein to albumin ratio OR C-reactive protein/albumin ratio OR C-reactive protein albumin ratio OR CRP/albumin ratio) AND (colorectal cancer OR rectal cancer OR colon cancer) AND (survival OR outcome OR mortality). The bibliographies of the included trials and related review articles were manually reviewed for potential missing studies. The protocol for this systematic review was registered on INPLASY (
https://inplasy.com/), and the registration number is INPLASY202140103. The protocol is available in full on the
inplasy.com (
https://doi.org/10.37766/inplasy2021.4.0103).
Inclusion and exclusion criteria
The inclusion criteria were as follows: (1) either retrospective or prospective studies reporting the association of pre-treatment CAR and the outcome of CRC patients, (2) participants having pathologically confirmed CRC, (3) CAR data before treatment, (4) include survival outcomes such as overall survival (OS) and/or disease-free survival (DFS)/progression-free survival (PFS) for analysis, (5) provide sufficient information for the extraction of the hazard ratio (HR) and associated 95% confidence intervals (CIs), and (6) studies published in English or Chinese. The exclusion criteria were as follows: (1) lack of sufficient data for extracting HR and the associated 95% CIs; (2) no evaluation of survival outcome; (3) use of post-treatment CAR data; (4) case reports, review articles, or conference abstracts; and (5) studies published in languages other than English or Chinese.
Data extraction and quality assessment
Two independent reviewers examined all retrieved articles and extracted the data using a predetermined form. The following information was extracted: the first author’s name, year of publication, country of study, tumor location, sample size, gender, mean age, cancer stage, type of treatment, the cut-off value of CAR, outcome measures, mean follow-up times, confounding factors of CAR, and analyzed models. Two reviewers evaluated the methodological quality of the enrolled studies independently, according to the Newcastle–Ottawa Quality Assessment Scale (NOS). The NOS contains nine items in three categories: participant selection (four items), comparability (two items), and exposure (three items). A study can be scored a maximum of 1 point for items in the selection and exposure domains and 2 points for the comparability domain [
22]. A study with NOS scores of 7 or higher was defined as a high-quality study. Any discrepancies between the two reviewers were resolved through discussion.
Statistical analysis
All statistical analyses were performed using Comprehensive Meta-Analysis software, version 3 (Biostat, Englewood, NJ, USA). The HRs and associated 95% CIs were directly extracted from the study while reporting survival analysis of OS or PFS/DFS and pooling the prognostic value of high pre-treatment CAR. The clinicopathological features associated with high CAR were also extracted from the studies if a study reported a correlation between CAR and clinicopathological factors. The heterogeneity between studies was determined using the Cochran Q-test and I2 statistics. I2 ≥ 50% indicated considerable heterogeneity. A random-effects model was employed to pool the HRs and ORs in this meta-analysis. Funnel plots, Begg’s test, and Egger’s test were used to examine potential publication bias. Statistical significance was defined as p-values < 0.05, except for the determination of publication bias, which was employed at p < 0.1. To evaluate the influence of publication bias, we used the trim-and-fill method. A sensitivity analysis was performed to confirm the robustness of the pooled results by removing each study individually. Moreover, subgroup analysis was performed to investigate the heterogeneity between eligible studies with the following features: study region (Western or Eastern), sample sizes (< 200 or ≥ 200), stage (stage I–III or stage IV), treatment methods (surgery, chemotherapy, or liver resection only), the cut-off value of CAR (< 0.1 or ≥ 0.1), and study method (multivariate or univariate analysis).
Discussion
We conducted this systematic review and meta-analysis with a total of 6329 participants to determine the prognostic value of CAR in CRC patients. This study included the most recent literature from both Eastern and Western countries. Our study showed that high pre-treatment CAR was associated with the poor survival, either overall or disease-free, of CRC patients. A double death risk is observed more in high pre-treatment CAR patients than in those with low pre-treatment CAR. Moreover, up to four-fold of the progression risk is observed in stage IV CRC patients receiving chemotherapy. The subgroup analysis revealed the constant effect of high pre-treatment CAR on survival, including stratification by region, sample size, stage, treatment type, or CAR cut-off value.
Additionally, a high pre-treatment CAR correlated with many advanced clinicopathological features, such as older age, advanced tumor size, stage, elevated CEA and NLR levels, and prominence more in the colon than in the rectum. In addition to these factors, a high level of CA-199 [
19] and an increased tumor depth [
23] were reported to be correlated with a high pre-treatment CAR. Consequently, CAR can be used as a prognostic biomarker for patients with CRC. Assessing CAR before treatment is essential not only for its prognostic value but also for stratifying high-risk patients.
It is well known that inflammation plays a critical role in tumorigenesis. Malignancies can trigger an intrinsic inflammatory response that affects the tumor cells’ microenvironment in tumor progression and metastasis [
36]. Therefore, several inflammation-related markers have been studied for use in cancer treatment. In recent decades, the neutrophil-to-lymphocyte ratio, platelet-to-lymphocyte ratio, GPS, and mGPS have been widely used to predict CRC outcomes [
6,
15,
16]. GPS and mGPS are composed of CRP and albumin, both of which may reflect the underlying inflammatory condition and nutritional status. Because only three scores characterize GPS and mGPS, most patients are categorized to have scores 0 or 1, typically considered to have a better prognosis than score 2. However, by using CAR, which has a quantitative nature with a continuous range of values, we can further divide the GPS/mGPS group with CAR. Ishizuka et al. found a superior overall survival in the low CAR subgroup than in the high CAR subgroup for CRC patients with GPS 0 and GPS 1, respectively [
19].
In addition to the predictive role of prognosis in patients with CRC, there are several CAR utilities in clinical practice. A multicenter study conducted by Hashimoto et al. found that high pre-operative CAR was an independent predictor of postoperative complications (HR 2.864,
p = 0.029) in patients aged 85 years or older who underwent primary tumor resection for CRC [
37].
Shibutani et al. found that the normalization of CAR at 8 weeks after initiating chemotherapy for stage IV CRC patients tended to have a better OS than the persistently high CAR subgroup [
23]. A study conducted by Zhou et al. also found that dynamic changes in inflammatory markers, including CAR, are associated with OS [
20]. Moreover, Tominaga et al. investigated 136 patients with stage III CRC receiving adjuvant chemotherapy and found a significant increase in the risk of severe side effects of chemotherapy (HR 7.06, 95% CI 2.51–19.88,
p < 0.01) in patients with high pre-treatment CAR [
30]. In addition to pre-treatment CAR usage, Ge et al. found that high postoperative CAR on POD-3 had a higher positive predictive value for postoperative complications than CRP alone [
38]. Another study conducted by Matsuoka et al. found the survival benefit of adjuvant chemotherapy in stage III CRC patients with high postoperative CAR after curative surgery but no superior outcome of adjuvant chemotherapy in the low postoperative CAR subgroup [
39].
As an acute-phase protein, CRP is mediated by many pro-inflammatory cytokines, such as IL-1, IL-6, and TNF-α [
12]. These cytokines suppress the synthesis of albumin under inflammatory conditions [
13,
14]. Thus, as a combination of these two proteins, CAR may reflect the severity of inflammation, which is believed to be correlated with tumor progression. This updated meta-analysis summarized the most recent studies and indicated the prognostic value of pre-treatment CAR for OS and DFS/PFS. Thus, CAR data should be used for both, checking the pre-treatment status of patients with CRC and other cancers, and observing the dynamic changes in CAR values during and post-treatment. More studies involving CAR are warranted to gain a complete understanding of its utility in CRC treatment.
This study has some limitations. First, all of the included studies were retrospective in design, the existence of co-morbidities influenced the CRP and albumin values, and selection bias potentially existed. Second, most of the included studies originated in the Eastern region. The pooled results may not be as consistent as they are in the Western region. Third, the CAR cut-off value varied between each study; therefore, we could not determine the most reliable clinical practice value. Fourth, we could only extract HRs and 95% CIs directly for the meta-analysis from each study. The existence of missing data may reduce the overall accuracy of the pooled results. Finally, publication bias did exist in this systemic review and may attributed to the following reasons. We only included the articles in English and Chinese although we did not set any limitation of language during the searching process. Unpublished papers with non-significant results may exist; however, there was no significant alteration in the pooled result after adding two missing studies in the funnel plot using the trim-and-fill analysis. We believe this meta-analysis is still reliable.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.