Conventional Laboratory Blood Indicators Are Valuable for Early Diagnosis of Colorectal Cancer

Objective: Some conventional laboratory indicators have been found to be of value for the diagnosis of colorectal cancer (CRC). The present study aimed to systematically analyze the diagnostic value of conventional laboratory blood indicators for CRC, especially for early CRC. Methods: A total of 505 patients with CRC (n=210), colorectal adenoma (CRA) (n=167) or polyp (CRP) (n=128) were retrospectively collected. Clinical, laboratory and imaging data available before treatment were extracted. The diagnostic performances of laboratory blood indicators for discriminating total and early CRCs from CRA and CRP (CRA&P) were evaluated. Results: Fifty-three of 76 (69.7%) laboratory blood indicators were signicant for discriminating CRC from CRA&P with areas under the receiver operating characteristic curve (AUC) ranging within 0.554-0.819, of these indicators, 17 had AUC > 0.7, three had AUC > 0.8, and ve had AUCs greater than that for carcinoembryonic antigen (CEA). Fifteen indicators had overall sensitivities comparable to CEA for the diagnosis of CRC (35.7-55.4% vs. 47.7%, all P>0.05) at a specicity of 90%, and they were not or weakly correlated with CEA (absolute r = 0.058-0.333). For differentiating early CRC (TNM stage I+II, n=102) from CRA&P, the sensitivities for the 15 indicators ranged within 30.4%-55.5% at a specicity of 90% and similar to stage III+IV CRC. Conclusion: Conventional laboratory blood indicators are valuable for early CRC diagnosis, and are comparable to or better than CEA.

been discovered and evaluated, including DNA and its transcription and epigenetics, mRNA and noncoding RNA, proteins and metabolites, these are far from being applied in clinical diagnosis [9].
Daily clinical and healthcare records are useful data for exploiting diagnostic indicators for tumors, including CRC [10]. It has been early observed that there is an association between complete blood count and colon cancer [11]. A recent meta-analysis revealed that the levels of red blood cell (RBC) count, hemoglobin (HGB), mean corpuscular volume (MCV), red blood cell distribution width (RBC-DW), white blood cell (WBC) count, and platelet (PLT) count are valuable for the diagnosis of CRC [12]. Apart from the PLT count, altered mean platelet volume (MPV) and platelet thrombocytocrit (PCT) might be valuable for the diagnosis and prognosis of CRC [13]. The indexes derived from the blood cell count, neutrophil-tolymphocyte ratio (NLR) and platelet-to-lymphocyte ratio (PLR), have been found to be useful in the diagnosis and early detection of CRC, especially in combination [14]. Apart from the metrics in the blood count test, several other laboratory indicators were found to have some diagnostic value for CRC, such as ecto-5'-nucleotidase (5'-NT) [15,16].
Indeed, some laboratory indicators are valuable for CRC diagnosis, to some extent, indicating that conventional laboratory data may be an alternative approach for CRC screening and detection. However, the value of clinical metrics for the diagnosis of CRC has not been systematically investigated. In the present study, the investigators retrospectively collected hospitalized patients with CRC, colorectal adenoma (CRA) and colorectal polyp (CRP), who underwent surgical or colonoscopic therapy, and systematically evaluated the diagnostic value of conventional blood test indicators for CRC, particularly highlighting the early diagnostic performances.

Demographic and laboratory characteristics of patients
A total of 505 patients were recruited for the present study, which included 210 CRC patients, 167 CRA patients, and 128 CRP patients. The demographic and laboratory blood data of these patients are shown in Table 1. A total of 76 blood indicators were analyzed, including blood cell analysis, biochemistry, tumor markers and coagulation test, and 49 (64.5%) of them signi cantly differed between CRC and CRA&P.

Pathological characteristics and clinical stages of colorectal cancer patients
The pathological characteristics and TNM stage of these CRC patients are presented in Table 2. Only 190 surgical patients had the TNM classi cation data available. The colorectal lesions were dominantly located in the colon (91.0%). Most of CRC were at TNM stage II or III (77.9%), with the size of 3-5cm (54.3%), histological type of adenocarcinoma (90%), and middle or poor differentiation (94.3%). The diagnostic value of each indicator for discriminating CRC from CRA + CRP (CRA&P) was evaluated using the ROC curve and 54 indicators with signi cant AUCs (all P < 0.05) are shown in Table 3. Among these indicators, 19 indicators had an AUC of > 0.7, three indicators had an AUC of > 0.8, and ve indicators had an AUC of greater than that for CEA.

The sensitivities of laboratory blood indicators for diagnosing various stages of CRC
The optimal cut-off value for each blood indicator for discriminating CRC from CRA&P was determined according to the ROC curve. Setting the speci city at 90%, the sensitivities of 16 clinical indicators with an AUC of ≥ 0.7 (ferritin and D-Dimer were excluded due to the small sample size) were calculated for the diagnosis of CRC, and compared with CEA and among the CRC stages. The overall sensitivity for the 16 indicators, except FIB, were comparable with CEA at a speci city of 90% for the diagnosis of CRC ( Fig. 2A, Table 4), and ve indicators presented higher overall sensitivities than that for CEA at 90% speci city, although the differences in sensitivity were not signi cant. The bivariate correlation analysis showed that there were no or weak correlations between these indicators and CEA (P < 0.001-0.274) (Fig. 2B).

The diagnostic value of laboratory blood indicators for early CRC
ROC curve analyses were performed to evaluate the early diagnostic value of conventional blood indicators, and eight indicators had AUCs greater than that for CEA, in terms of differentiating early CRC (stage I + II) from CRA&P (Fig. 3).

Discussion
The present study retrospectively analyzed the difference in conventional laboratory blood metrics  [19]. The iron de ciency-related indicator, ferritin, also signi cantly decreased in CRC and this is consistent to a previous report [20], which further con rms the above anemia-related results. In addition, these present results exhibited that the levels of hemoglobin-related indicators (HCT, MCH, MCHC) were signi cantly reduced in the CRC of stage I compared with the CRA&P (data not shown), but there were no signi cant differences among the various stages (Table 4). These are similar to the results of a recent study, in which the levels of HGB, MCV and serum ferritin (SF) decreased shortly before the CRC diagnosis [21].
Apart from the RBC-related indicators, most of the WBC-and PLT-related indicators were valuable for the CRC diagnosis, but these were not as good as the RBC-related indicators, and only two indicators (NEUT% and LYMP%) had an AUC of > 0.7. WBC-related indicators indicate the in ammatory condition of body. In ammation has been well-known to be closely associated with the onset and progression of cancer, including CRC. The in ammatory cells and cytokines in tumors have been considered to more likely contribute to tumor growth, progression, and immunosuppression, when compared to mounting an effective host anti-tumor response [22]. In addition to the lower circulation lymphocyte number and percentage, and higher circulation neutrophil number and percentage in CRC, when compared to CRA&P, signi cantly higher levels of C-reactive protein (C-RP) and brinogen (FIB) were also found in the present study, and both of these presented AUCs greater than 0.7 for the diagnosis of CRC. These results are consistent with the reports, in which low tumor CD4 + T-lymphocyte in ltration is associated with elevated C-RP concentration and poor cancer-speci c survival in CRC patients [23] and FIB is epidemiologically and mechanistically linked with diseases with an in ammatory component [24].
Furthermore, 10 of 20 liver function indicators with AUCs > 0.6 for discriminating CRC from CRA&P, PALB, RBP and CHE presented the greatest AUCs (0.807-0.819) in all indicators, with the diagnostic performance at the "good" level. This was superior to CEA, and but was not different between early-and late-stage CRC patients. Although it has been early found that blood levels of PALB and RBP are correlated to the nutrition and prognosis of CRC [25], the diagnostic value of PALB has just been recently reported. Sun et al. [26] used the ratio of circulating FIB to PALB levels to diagnose CRC, and obtained an AUC of 0.845. In the present study, PALB and RBP had a similar diagnostic performance for CRC, but PALB had a higher sensitivity than ALB. This may be because PALB has a much shorter half-life than ALB (2 vs. 20 days), as well as a smaller body pool and a more rapid synthesis rate. Therefore, PALB was considered the most sensitive and stable indicator, when compared to ALB, in terms of nutritional evaluation [27,28]. Malnutrition in CRC is more frequent, when compared to other common cancers [29], and early-stage CRC can present apparent sarcopenia, and correlate to survival [30]. These features make PALB valuable in the early diagnosis of CRC. RBP strongly interacts with PALB, and circulates in plasma in a 1:1 molar RBP-PALB complex [31]. Thus, a similar diagnostic performance was observed in the present study.
Among all indicators, serum cholinesterase (CHE) exhibited the highest AUC for the diagnosis of CRC, including early CRC. The reduced serum CHE activity has been early reported in cancer, when compared to normal control [32], and in CRC, when compared to non-cancer patients [33]. Furthermore, this has been considered to be a prognostic factor for CRC patients [34], but this has not been evaluated for the diagnosis of CRC. The diagnosis value of CHE for CRC is probably correlated to its association with in ammation and malnutrition. It was found that serum CHE activity inversely correlates with subclinical in ammation [35] and severe systemic in ammation [36]. In the present study, negative correlations were observed between serum CHE activity and C-reactive protein (r = -0.278) and NEUT% (r = -0.275). CHE activity reduction in in ammation is correlated to the cholinergic anti-in ammatory pathway, in which acetylcholine, the substrate of CHE, plays an anti-in ammatory function and regulates the CHE activity in a negative-feedback manner [36]. Low CHE activity is also a serum marker of nutritional status in patients with CRC [37], and an increase of CHE activity was found after nutritional support therapy [38]. In addition, it was found that CHE is downregulated in CRC tissues, when compared to paired normal tissues, and it was presumed that the over-stimulating muscarinic receptors via increasing acetylcholine is correlated to the gut carcinogenesis [39]. These above ndings provide the rationale for serum CHE as a valuable biomarker for CRC diagnosis.
CEA, a classical biomarker of CRC, presented fair and similar diagnostic performances for overall CRC and early CRC (AUC = 0.758 and 0.742) in the present study. CEA has been recommended to be used in CRC relapse monitoring [40]. However, its low sensitivity limits its application in early diagnosis. In the present study, eight indicators (HGB, HCT, PALB, ALB, RBP, CHE, Ferritin and B2MG) had greater AUCs, when compared to CEA, for the diagnosis of early CRC. Furthermore, these had null or weak correlations with CEA, indicating that these indicators are superior or at least equal to CEA for the early diagnosis of CRC, in terms of diagnostic performance. In the eight indicators, beta-2-microglobulin (B2MG) is a biomarker for kidney ltration and cell turnover. This has been found to be elevated in some cancers, including CRC [41], and negatively correlated to the prognosis of recurrent CRC [42]. In a population-based cohort study followed-up for a maximum of 17 years, participants with the highest quartile of serum Β2MG concentration had a 121% higher risk of CRC incidence, when compared to those with the lowest quartile, and furthermore, this was much higher than the risk of total cancer incidence (25%) and independence of conventional clinical factors [43], indicating that B2MG is strongly associated with CRC incidence risk. The mechanism for the association of serum Β2MG concentrations with CRC carcinogenesis remains unclear. This is probably correlated to the pro-angiogenic, pro-tumorigenic, driving innate pro-in ammatory cytokines and growth promoting factors, epithelial-mesenchymal transition, and cell turnover [43]. For ferritin, which is a well-known iron binding protein, the reduction in serum ferritin level could be prior to anemia [20]. This is more remarkable in eastern countries, when compared to western countries, according to a meta-analysis [44].
Although the diagnostic value of conventional laboratory blood indicators for CRC was systematically and comprehensively analyzed in the present study, there were several limitations. First, we did not perform a multivariate analysis on these indicators due to the missing values in some indicators.
Therefore, the independent indicators for CRC diagnosis need to be clari ed in future studies. Second, since systematical blood tests could not be performed in outpatients and normal controls, merely inpatients with CRC, CRA and CRP were enrolled in the present study, which may cause the results to be inconsistent with the situation in the real world. Third, as a monocenter and retrospective study, further studies with a prospective and multicenter design and multivariate analysis are warranted to elucidate and validate the diagnostic signi cance of conventional laboratory blood tests in patients with CRC.
In summary, the investigators retrospectively and systematically analyzed the diagnostic performance of conventional laboratory blood indicators in differentiating CRC and CRA&P. It was found that most of the indicators have certain value for the diagnosis of CRC, including early CRC. Indicators correlated to anemia, nutrition status, and in ammation had greater value, when compared to the other indicators, and some of these were superior to that of CEA. Prospective studies with a more rigid design should be performed to validate these present results.

Patients
Patients with CRC, CRA, or CRP, who were hospitalized for surgical or colonoscopic therapy in the Three A liated Hospital of Nanchang University from March 2014 to July 2019, were collected. The medical records of these patients were reviewed, and the clinical data available were extracted, including the demographics, medical history, laboratory blood tests (blood cell analysis, biochemistry, coagulation function, tumor markers, etc.), pathological diagnostic data, medical image examination (ultrasonography, and CT and/or MRI), and therapy data (surgery, colonoscopy and other treatments).
The laboratory data collected were the results of the rst test after admission prior to therapy, which was close to the time of diagnosis.
Age at the laboratory test was recorded. All colorectal lesions were pathologically con rmed. These CRC patients were staged according to the 8th edition of the American Joint Committee on Cancer (AJCC) TNM staging system [17]. This study was approved by the Ethical Committee of the First A liated Hospital of Nanchang University, which also waived the need for informed consent for the retrospective study. All methods were performed in accordance with the relevant guidelines and regulations, including the Declaration of Helsinki.
Patients with one of following conditions were excluded: (1) patients with no de nite pathological diagnosis; (2) patients with recurrent or secondary colorectal cancer; (3) patients with other concurrent cancers; (4) patients suffering from liver diseases, hematological diseases, or other diseases that can affect the results of the laboratory blood tests; (5) patients who received anti-tumor therapies before surgery; (6) patients who received special therapies that affect the results of the laboratory blood tests, such as iron agent, anticoagulant, anti-lipemic agent, nonsteroidal anti-in ammatory drugs, and blood transfusion; (7) patients with incomplete data for the TNM staging of CRC or the de nite diagnosis. Figure 1 shows the enrollment of patients.

Statistical analysis
The patient demographics, laboratory tests, and clinical and pathological characteristics were descriptively summarized. Continuous data were expressed as mean ± standard deviation (SD), and enumeration data were expressed in frequency and percentage. The difference of each indicator among the three groups was compared using ANOVA or Pearson's Chi-squared test, according to the type of variable. The difference of indicators between TNM stages was compared using Pearson's Chi-squared test or the Fisher's exact test. The area under the receiver operating characteristic (ROC) curve (AUC) was utilized to evaluate the diagnostic performance of each indicator. The McNemar test was used to compare the diagnostic sensitivity between CEA and other laboratory indicators. The Pearson's correlation was used to analyze the bivariate correlation between the blood indicators. A two-sided Pvalue of < 0.05 was considered statistically signi cant. The statistical analyses were carried out using the SPSS version 24.0 software (IBM, NY, USA).

Figure 1
The owchart of patient enrollment.
Page 26/27  Table 1 for the abbreviations.