The impact of chemotherapy-associated hemoglobin on prognosis of colorectal cancer patients receiving adjuvant chemotherapy

Abstract

BACKGROUND AND OBJECTIVE:

The association of chemotherapy-associated hemoglobin and survival of colorectal cancer (CRC) receiving adjuvant chemotherapy is uncertain. We sought to explore the prognostic value of chemotherapy-associated hemoglobin in CRC receiving adjuvant chemotherapy and the best cut point affecting prognosis.

METHODS:

Three hundred and twenty stage II and III CRC patients receiving adjuvant FOLFOX chemotherapy from March 2003 to March 2012 were enrolled. The associations between chemotherapy-associated hemoglobin (the absolute levels of post-chemotherapy) or chemotherapy-associated hemoglobin change (change between the pre- and post-chemotherapy hemoglobins) and disease free survival (DFS) or overall survival (OS) of CRC, and the best cut point were investigated.

RESULTS:

Log rank test showed the best cut points for chemotherapy-associated hemoglobin and chemotherapy-associated hemoglobin change were respectively 90 g/L, 30 g/L. Cox regression model showed chemotherapy-associated hemoglobin $<$ 90 g/L was the independent prognostic factor for DFS (HR, 2.221; 95% CI $=$ 1.157–4.262), OS (HR, 2.058; 95% CI $=$ 1.009–4.197), respectively, but no association of chemotherapy-associated hemoglobin change $⩾$ 30g/L and DFS (HR, 2.063; 95% CI $=$ 0.929–4.583), OS (HR, 1.386; 95% CI $=$ 0.553–3.471) was found.

CONCLUSIONS:

Chemotherapy-associated hemoglobin $<$ 90 g/L has a significant prognostic value in CRC receiving adjuvant chemotherapy, which is a significant biomarker in the individualized management and may suggest the simple indication for the treatment of anemia in adjuvant chemotherapy in CRC.

1. Introduction

Combination of fluorouracil (5-Fu), leucovorin (LV) and oxaliplatin chemotherapy (FOLFOX), which has been widely accepted as the standard adjuvant chemot- herapy for stage III and stage II colorectal cancer (CRC), has reduced the risk of relapse after surgery and improved survival significantly [1]. However, the responses to chemotherapy vary in different subpopulations. Thus, approximately 20%–40% patients receiving adjuvant chemotherapy after a curative operation experience recurrence [2, 3, 4, 5]. It is important to identify the risk factors which impact recurrence and survival of CRC patients receiving adjuvant chemotherapy.

As a frequent finding, anemia occurs in 18%–70% of cancer patients [6, 7, 8]. In Chinese population, the normal hemoglobin levels are 120–160 g/L in male, 110–150 g/L in female, respectively [9]. Chemotherapy-induced anemia, which is a common complication in chemotherapy, occurs in 30%–70% of cancer patients receiving chemotherapy [10, 11, 12]. The National Cancer Institute Common Terminology Criteria for Adverse Events version 4.0 (NCI-CTCAE 4.0) for the assessment of therapy induced anemia including chemotherapy-induced anemia in CRC included the following (https://ctep.cancer.gov/protocolDevelopm-ent/electronic_applications/ctc.htm): grade 1 (mild), 100 g/L to within normal limits; grade 2 (moderate), 80–100 g/L; grade 3 (severe), 65–80 g/L; and grade 4 (life threatening), less than 65 g/L. Anemia contributes to tumor progression in different types of cancer [7, 13, 14, 15]. Most of the reports focused on the impact of preoperative anemia on cancer, but the association of chemotherapy-associated hemoglobin and survival of colorectal cancer receiving adjuvant chemotherapy is uncertain. Herein, we investigated the prognostic relationship between chemotherapy-associated hemoglobin and survival of CRC patients receiving adjuvant FOLFOX chemotherapy. Moreover, we selected the best cut point of chemotherapy-associated hemoglobin which impacted the clinical outcome to guide the individualized management in CRC receiving chemotherapy.

2. Materials and methods

2.1 Patient selection

Stage II and III pathology-proven CRC patients from March 2003 to March 2012 in the Second Affiliated Hospital of Guangzhou Medical University, who received adjuvant FOLFOX chemotherapy, were enrolled in our retrospective study. The deadline of follow up was April 2013. Stage II CRC patients receiving adjuvant FOLFOX chemotherapy only included those who had at least one of the following high risk factors: T4, tumor perforation, bowel obstruction, poorly differentiated tumor, venous invasion, or less than 10 lymph nodes examined, or those who had the intention to receive chemotherapy when they were informed of the efficacy and risk of chemotherapy. The exclusion criteria included the following: anemia induced by other factors such as hematologic diseases, chronic renal diseases, folate deficiency; primary prophylactic treatment of iron supplements, transfusion, or erythropoiesis-stimulating agents (ESAs) administration following chemotherapy. Other inclusion and exclusion criteria were defined in our previous report [15] – inclusion criteria: At least 3 cycles of adjuvant chemotherapy, no tumor recurrence during chemotherapy, WHO performance status (PS) 0–1, adequate pretreatment renal (pretreatment creatinine clearance $⩾$ 60 mL/min), and hepatic functions (pretreatment bilirubin $⩽$ 1.5 upper limit of normal, pretreatment alanine aminotransferase and/or aspartate aminotransferase $⩽$ 2.5 upper limit of normal), adequate baseline bone marrow (absolute baseline neutrophil counts $⩾$ 2.0 $\times$ 10 ${}^9$ cells/L, absolute baseline lymphocyte counts $⩾$ 1.0 $\times$ 10 ${}^9$ cells/L, baseline platelet counts $⩾$ 100 $\times$ 10 ${}^9$ cells/L); exclusion criteria: biologic or immunotherapy, concomitant or neoadjuvant radiotherapy, previous systemic chemotherapy or neoadjuvant chemotherapy, primary prophylactic administration of granulocyte colony-stimulating factor (G-CSF) following chemotherapy, previous malignancies other than colorectal cancer, documented human immunosuppression.

2.2 Ethics statement

The study conforms with the principles expressed in the Declaration of Helsinki. The study was approved by the institutional review boards of the Second Affiliated Hospital of Guangzhou Medical University and the institutional review boards of the Second Affiliated Hospital of Nanchang University. All participants signed an informed consent approved by the institutional Review Boards.

Figure 1.

Hemoglobin levels at baseline and after chemotherapy.

2.3 FOLFOX treatment and data collection

The FOLFOX regimen consisted of a 2-h intrav- enous infusion of 85 mg/m ${}^2$ oxaliplatin and 400 mg/m ${}^2$ LV, followed by 400 mg/m ${}^2$ bolus 5-FU plus a 46-h intravenous infusion of 2400 mg/m ${}^2$ 5-FU, repeated every 2 weeks. Chemotherapy was delayed due to severe toxicity and the doses of oxaliplatin and 5-FU were reduced by 15% in subsequent cycles. Chemotherapy was discontinued due to unacceptable toxicity.

From the medical records we collected the data of pretreatment albumin, pretreatment carcinoembryonic antigen (CEA), differentiation, sex, age, location, stage, lymphocyte, hemoglobin. The evaluation of WHO PS, lymphocyte and hemoglobin test was performed before each next chemotherapy cycle. The nadir lymphocytes, hemoglobin after chemotherapy were recorded in our study. Chemotherapy-associated hemoglobin $=$ The absolute levels of nadir hemoglobin after chemotherapy. The chemotherapy-associated he- moglobin change $=$ the baseline hemoglobin prior to chemotherapy minus the absolute levels of nadir hemoglobin after chemotherapy. Albumin, CEA, and stage were defined as our previous report [16]: Albumin was divided into the following two groups: $⩾$ or $<$ 35 g/L, CEA was divided into the following two groups: $⩾$ or $<$ 10 ng ml ${}^{-1}$ , staging was performed according to the American Joint Committee on Cancer (AJCC, seventh edition). Our previous report has shown chemotherapy-associated lymphopenia $<$ 0.66 $\times$ 10 ${}^9$ /L has a significant impact on disease free survival (DFS), chemotherapy-associated lymphopenia $<$ 0.91 $\times$ 10 ${}^9$ /L has a significant impact on overall survival (OS) of CRC receiving adjuvant chemotherapy, respectively [16]. Thus, we included chemotherapy-associated lymphocytes in our study. In the Cox proportional hazard model, chemotherapy-associated lymphocytes were divided into $⩾$ or $<$ 0.66 $\times$ 10 ${}^9$ /L for DFS, $⩾$ or $<$ 0.91 $\times$ 10 ${}^9$ /L for OS, respectively.

2.4 Statistical analysis

Log rank test and life table were used to estimate DFS and OS. Minimum $p$ -value approach, or alternatively the maximum statistic approach based on the log rank test statistic, proposed by Williams et al. [17], Altman et al. [18] and Contal and O’Quigley [19], was used to estimate the best cut point of chemotherapy-associated hemoglobin or the chemotherapy-associated hemoglobin change affecting DFS or OS. Multivariate analysis and Cox proportional hazard models were used to determine the independent impact of chemotherapy-associated hemoglo- bin on DFS or OS. All statistical analyses were performed by Statistical Package of Social Sciences 19.0 software and 2-sided. $P$ value $<$ 0.05 was considered to be statistically significant.

Table 1

Log rank statistic and $p$ value for the potential cut points of chemotherapy-associated hemoglobin evaluating disease free survival (DFS) or overall survival (OS) of colorectal cancer

Potential	DFS		OS
cut points	Log rank statistic	$P$ value	Log rank statistic	$P$ value
150 g/L	0.443	0.506	0.468	0.494
140 g/L	0.332	0.564	0.164	0.685
130 g/L	1.060	0.303	1.712	0.191
120 g/L	2.623	0.105	0.748	0.387
110 g/L	6.584	0.010	2.623	0.105
100 g/L	3.251	0.071	4.163	0.041
90 g/L	22.568	$<$ 0.0001	15.127	$<$ 0.0001
80 g/L	13.149	$<$ 0.0001	9.336	0.002
70 g/L	8.066	0.005	1.028	0.311
50 g/L	19.358	$<$ 0.0001	37.029	$<$ 0.0001
40 g/L	6.946	0.008	11.239	0.001

Table 2

Life table for chemotherapy-associated hemoglobin $⩾$ 90 g/L

Interval start	DFS			OS
time (months)	Number exposed to risk	Number of terminal events	Cumulative proportion surviving at end of interval	Number exposed to risk	Number of terminal events	Cumulative proportion surviving at end of interval
0	265.5	21	92%	266	6	98%
12	221.5	21	83%	236	16	91%
24	151.5	7	80%	169.5	13	84%
36	105	6	75%	114	9	77%
48	80.5	3	72%	82.5	2	76%
60	67	2	70%	69	2	73%
72	55.5	0	70%	57.5	1	72%
84	43	0	70%	44	1	71%
96	26	0	70%	26.5	1	68%
108	10	2	56%	9	0	68%

Table 3

Life table for chemotherapy-associated hemoglobin $<$ 90 g/L

Interval start	DFS			OS
time (months)	Number exposed to risk	Number of terminal events	Cumulative proportion surviving at end of interval	Number exposed to risk	Number of terminal events	Cumulative proportion surviving at end of interval
0	54	7	87%	54	2	96%
12	43.5	14	59%	48	6	84%
24	22	4	48%	33	8	64%
36	12.5	2	41%	17.5	4	49%
48	7	0	41%	8.5	1	43%
60	4.5	1	32%	4.5	1	34%
72	2.5	0	32%	2.5	0	34%
84	2	0	32%	2	0	34%
96	2	0	32%	2	0	34%
108	1	0	32%	1	0	34%

3. Results

3.1 Characteristics of the study population

320 CRC patients who fulfilled the eligibility and exclusion criteria were included in the present study with a median follow-up time of 33 months (range 7–120 months), a median age of 59 years (a range of 22–82 years). Of the 320 cases, 33 cases had to stop adjuvant chemotherapy for severe toxicity. Of the 320 cases, 177 (55.3%) were male and 143 (44.7%) were female. One hundred and forty-three (44.7%) cases were rectal cancer, 99 (30.9%) cases were left colon cancer, and 78 (24.4%) were right colon cancer. According to the American Joint Committee on Cancer (AJCC, seventh edition), there were 185 (57.8%) stage II cases and 135 (42.2%) stage III cases. The population in our study had a mean baseline hemoglobin of 112.2 $\pm$ 16.7 g/L (mean $\pm$ standard deviation), a mean chemotherapy-associated hemoglobin (nadir hemoglobin after chemotherapy) of 104.8 $\pm$ 16.5 g/L (mean $\pm$ standard deviation), and a mean hemoglobin change of 7.4 $\pm$ 12.8 g/L (mean $\pm$ standard deviation) during chemotherapy. The actual levels of baseline hemoglobin before chemotherapy and chemotherapy-associated hemoglobin after chemotherapy were shown in Fig. 1. According to NCI-CTCAE 4.0, grade 3/4 anemia was observed in 20 (6.25%) cases.

Table 4

Log rank statistic and $p$ value for the potential cut points of chemotherapy-associated hemoglobin change evaluating DFS or OS of colorectal cancer

Potential	DFS		OS
cut points	Log rank statistic	$P$ value	Log rank statistic	$P$ value
$-$ 20 g/L	0.321	0.571	0.468	0.494
$-$ 10 g/L	0.000	0.984	0.515	0.473
0 g/L	0.014	0.907	0.509	0.476
10 g/L	2.852	0.091	2.461	0.117
20 g/L	8.309	0.004	2.078	0.149
30 g/L	16.563	$<$ 0.0001	5.254	0.022
40 g/L	6.939	0.008	1.137	0.286
50 g/L	2.066	0.151	0.050	0.823
60 g/L	0.372	0.542	0.403	0.526

3.2 The best cut point of chemotherapy-associated hemoglobin, chemotherapy-associated hemoglobin change on CRC survival

Table 5

Prognostic value of chemotherapy-associated hemoglobin $<$ 90 g/L, chemotherapy-associated hemoglobin change $⩾$ 30 g/L for DFS or OS of colorectal cancer

Variables	DFS			OS
	Hazard ratio	95% confidence intervals	$P$ -value	Hazard ratio	95% confidence intervals	$P$ -value
Pretreatment albumin
$⩾$ 35 g/L	1.00 (ref.)			1.00 (ref.)
$<$ 35 g/L	0.792	0.470–1. 334	0.380	1.160	0.684–1.966	0.582
Pretreatment CEA ${}^{\mathrm{a}}$
$<$ 10 ng ml ${}^{-1}$	1.00 (ref.)			1.00 (ref.)
$⩾$ 10 ng ml ${}^{-1}$	2.437	1.538–3.863	$<$ 0.0001	2.759	1.649–4.618	$<$ 0.0001
Differentiation
Well	1.00 (ref.)			1.00 (ref.)
Moderately	1.310	0.618–2.777	0.481	1.124	0.475–2.661	0.791
Low	0.842	0.297–2.385	0.745	1.121	0.360–3.494	0.844
Sex
Male	1.00 (ref.)			1.00 (ref.)
Female	1.362	0.874–2.122	0.172	1.295	0.796–2.109	0.298
Age
$⩽$ 49 years	1.00 (ref.)			1.00 (ref.)
50–60 yeas	1.378	0.750–2.531	0.302	1.247	0.611–2.543	0.544
$>$ 60 years	1.285	0.734–2.249	0.380	1.492	0.783–2.845	0.224
Location
Rectum	1.00 (ref.)			1.00 (ref.)
Left colon cancer	0.721	0.427–1.219	0.222	0.872	0.490–1.565	0.643
Right colon cancer	0.845	0.498–1.432	0.531	0.783	0.426–1.440	0.431
Stage
II	1.00 (ref.)			1.00 (ref.)
III	2.729	1.723–4.323	$<$ 0.0001	2.809	1.668–4.729	$<$ 0.0001
Chemotherapy-associated lymphocytes
$⩾$ 0.66 $\times$ 10 ${}^9$ /L (DFS)/	1.00 (ref.)
$⩾$ 0.91 $\times$ 10 ${}^9$ /L (OS)
$<$ 0.66 $\times$ 10 ${}^9$ /L (DFS)/	1.593	0.894–2.839	0.114	1.557	0.944–2.568	0.083
$<$ 0.91 $\times$ 10 ${}^9$ /L (OS)
Baseline hemoglobin	1.004	0.987–1.022	0.613	1.006	0.987–1.025	0.549
Chemotherapy-associated hemoglobin change
$<$ 30 g/L	1.00 (ref.)			1.00 (ref.)
$⩾$ 30 g/L	2.063	0.929–4.583	0.075	1.386	0.553–3.471	0.486
Chemotherapy-associated hemoglobin
hemoglobin $⩾$ 90 g/L	1.00 (ref.)			1.00 (ref.)
hemoglobin $<$ 90 g/L	2.221	1.157–4.262	0.016	2.058	1.009–4.197	0.047

${}^{\mathrm{a}}$ Carcinoembryonic antigen.

Our study had a range of 38–155 g/L for chemoth- erapy-associated hemoglobin. There were 11 distinct potential cut points: 150 g/L, 140 g/L, 130 g/L, 120 g/L, 110 g/L, 100 g/L, 90 g/L, 80 g/L, 70 g/L, 50 g/L, 40 g/L. The maximum value of log rank test statistic occurred at 90 g/L for DFS, 50 g/L for OS, respectively (Table 1). However, 50 g/L is an extreme cut point, as only 2 cases suffered from chemotherapy-associated hemoglobin $<$ 50 g/L. If we selected 50 g/L as the cut point, we achieved the study power of 0.311 (two-sided test, $\alpha =$ 0.05). Therefore 50 g/L is an improper cut point. The following maximum value of log rank test statistic occurred at 90 g/L for OS (Table 1). Thus we selected 90 g/L as the best cut point of chemotherapy-associated hemoglobin affecting DFS, OS with a $p$ value $<$ 0.0001 (Table 1). The number of events, number at risk at relevant time points and cumulative proportion surviving at end of interval were shown in Tables 2 and 3. Our study had a range of $-$ 26–69 g/L for chemotherapy-associated hemoglobin change. There were 9 distinct potential cut points: $-$ 20 g/L, $-$ 10 g/L, 0 g/L, 10 g/L, 20 g/L, 30 g/L, 40 g/L, 50 g/L, 60 g/L. The maximum value of log rank test statistic occurred at 30 g/L for DFS, OS with a $p$ value $<$ 0.0001, respectively (Table 4). Thus we selected 30 g/L as the best cut point of chemotherapy-associated hemoglobin change affecting DFS, OS.

3.3 The prognostic value of chemotherapy- associated hemoglobin $<$ 90 g/L, chemotherapy- associated hemoglobin change $⩾$ 30 g/L for colorectal cancer

Figure 2.

DFS or OS curve by Kaplan-Meier method. (A) Chemotherapy-associated hemoglobin $<$ 90 g/L was associated with shorter DFS ( $P<$ 0.0001). (B) Chemotherapy-associated hemoglobin $<$ 90 g/L was associated with shorter OS ( $P<$ 0.0001).

Figure 3.

DFS or OS curve of stage II colorectal cancer by Kaplan-Meier method. (A) Chemotherapy-associated hemoglobin $<$ 90 g/L was associated with shorter DFS of stage II colorectal cancer ( $P=$ 0.004). (B) Chemotherapy-associated hemoglobin $<$ 90 g/L was associated with shorter OS of stage II colorectal cancer ( $P=$ 0.039).

Figure 4.

DFS or OS curve of stage III colorectal cancer by Kaplan-Meier method. (A) Chemotherapy-associated hemoglobin $<$ 90 g/L was associated with shorter DFS of stage II colorectal cancer ( $P=$ 0.002). (B) Chemotherapy-associated hemoglobin $<$ 90 g/L was associated with shorter OS of stage II colorectal cancer ( $P=$ 0.008).

To control the possible confounding of the main effects of chemotherapy-associated hemoglobin $<$ 90 g/L, chemotherapy-associated hemoglobin change $⩾$ 30 g/L on DFS/OS, respectively, the clinicopathological factors (pretreatment albumin, pretreatment CEA, differentiation, sex, age, location, stage, baseline hemoglobin and chemotherapy-associated lymphocytes) were further adjusted for in the multivariate cox regression model. As shown in Table 5, cox model showed no association of chemotherapy-associated hemoglobin change $⩾$ 30 g/L and DFS (HR, 2.063; 95% CI $=$ 0.929–4.583), OS (HR, 1.386; 95% CI $=$ 0.553–3.471). Moreover, cox regression model showed no association of baseline hemoglobin (pre-chemotherapy hemoglobin levels) and DFS (HR, 1.004; 95% CI $=$ 0.987–1.022), OS (HR, 1.006; 95% CI $=$ 0.987–1.025). However, cox regression model showed chem- otherapy-associated hemoglobin $<$ 90 g/L remained the independent prognostic factor for DFS (HR, 2.221; 95% CI $=$ 1.157–4.262), OS (HR, 2.058; 95% CI $=$ 1.009–4.197). However, Kaplan-Meier method showed chemotherapy-associated hemoglobin $<$ 90 g/L was associated with shorter DFS ( $P<$ 0.0001) and OS ( $P<$ 0.0001) of stage II and III CRC (Fig. 2). Moreover, Kaplan-Meier method showed chemotherapy-associated hemoglobin $<$ 90 g/L was associated with shorter DFS ( $P=$ 0.004) and OS ( $P=$ 0.039) of stage II CRC (Fig. 3), shorter DFS ( $P=$ 0.002) and OS ( $P=$ 0.008) of stage III CRC (Fig. 4), respectively.

4. Discussion

Anemia may lead to the more progressive phenotype and the worse survival in malignancy. Preoperative anemia has been proved to be associated with the advanced stage and the worse survival in endometrial carcinomas [7]. Similarly, preoperative anemia is associated with lymph node metastasis, lymphovascular invasion, recurrence-free and cancer-specific survival in upper tract urothelial carcinoma [13]. Moreover, the impacts of pretreatment hemoglobin on progression, survival have been found in advanced esophageal cancer [14], breast cancer [20], hepatocellular carcinoma [20], nasopharyngeal carcinoma [20] and esoph- ageal carcinoma [20]. However, the data about the association of hemoglobin variation induced by myelosuppression therapy and survival in cancer are few. In our study, the multivariate cox regression model showed the risk effect of chemotherapy-associated hemoglobin on survival of CRC was independent of pretreatment hemoglobin, chemotherapy-associated lymphopenia. To the best of our knowledge, our study, for the first time, reported the association of chemotherapy-associated hemoglobin and survival of CRC receiving adjuvant chemotherapy. Our study indicates that while chemotherapy improves survival of CRC, the hemoglobin $<$ 90 g/L induced by chemotherapy may drive recurrence or metastasis of CRC, which may partly explain the failure of adjuvant chemotherapy in CRC.

The mechanism for the association of chemothera- py-associated hemoglobin and survival of CRC remains unclear. Anemic patients suffer from poorer tissue oxygenation [21, 22]. Hypoxic microenvironment under anemic condition may be responsible for the tumor recurrence or chemoresistance. Molecular changes in response to hypoxia, such as hypoxia-inducible factors (HIFs) family expression, contribute to carcinogenesis, tumor progression, chemoresistance or radioresistance. Hypoxia activates HIF-1alpha/G-protein estrogen receptor (GPER) signaling pathway involved in VEGF expression, which plays an important role in the formation of tumor microenvironment towards cancer progression [23]. HIF-1 $\alpha$ drives hypoxia-induced multidrug resistance (MDR) to chemotherapeutic drugs in laryngeal carcinoma cells [24]. Moreover, Snail1 gene activation regulated by HIF-2 $\alpha$ contributes to the effect of hypoxia on melanoma progression [25]. Besides HIFs, FGFR3 expression and OCT4 reduction induced by hypoxia may play the similar role in malignancy [26, 27]. Moreover, the effect of chemotherapy-associated hemoglobin on survival of CRC may be partly explained by the suppression of erythropoiesis leading to the increased hepcidin. Suppression of erythropoiesis increases hepcidin level, the hemolysis and anemia decrease hepcidin expression only when erythropoiesis is functional [28]. The elevated hepcidin might be correlated with carcinogenesis, development of tumor, a metastatic tumor and the poor prognosis [29, 30, 31]. The molecular changes mentioned above may contribute partly to the findings in our study.

According to the National Cancer Institute (NCI) and the World Health Organization (WHO) toxicity criteria, the classification of more severe grades of anemia is the same, but the classification of lesser grades differs slightly [32]. However, based on the findings by Vaupel, hemoglobin levels of less than 120 g/L results in significant tumor hypoxia altering its behavior [33]. Moreover, 120 g/L was used as the cut point of anemia in cancer in some reports [34, 35]. However, when we chose the cut point of 120 g/L in the present study, no impact of chemotherapy-associated hemoglobin $<$ 120 g/L on the survival of CRC was found(data not shown). Our study showed the best cut point of chemotherapy-associated hemoglobin was 90 g/L for DFS, OS and Cox regression model confirmed the prognostic value of chemotherapy-associated hemog- lobin $<$ 90 g/L for DFS, OS. Thus, it is reasonable that 90 g/L may be the reliable cut point affecting worse DFS/OS, respectively, for stage II and III CRC receiving adjuvant chemotherapy.

The hemoglobin cut point defined by ROC analysis has important clinical significance in adjuvant CRC chemotherapy. First, CRC cases undergoing FOLFOX regimen receive the standard dose according to the body surface area, however, under the given dose, the hemoglobin change response to adjuvant chemotherapy differs in different subpopulations. The dose will be reduced only if the patients suffer from severe toxicity such as III grade or more severe anemia (hemoglobin $<$ 80 g/L). The cut point of chemothe- rapy-associated hemoglobin $<$ 90 g/L selected in our study may provide another indication to reduce the dose of FOLFOX regimen for CRC. Thus it is significant in the individualized treatment of CRC chemotherapy. Second, the intervention threshold of hemoglobin in malignancy varied [36, 37], thus the cut point in our study suggest the simple indication for the treatment of anemia. Those who have chemotherapy-associated hemoglobin below the cut point should undergo active management to remain the hemoglobin above the cut point.

We propose chemotherapy-associated hemoglobin $<$ 90 g/L, but not baseline hemoglobin, or chemothe- rapy-associated hemoglobin change, as the potential prognostic factor for CRC receiving adjuvant chemo- therapy. Hemoglobin $<$ 90 g/L induced by chemotherapy may play a more important role than pretreatment hemoglobin in the prognosis of CRC receiving adjuvant chemotherapy. Moreover, our findings extend the clinical significance of chemotherapy-associated hemoglobin by analyzing the cut points to guide the individualized adjuvant chemotherapy in CRC. However, the sample size was relatively small in the present study. We await the larger, multicenter study to prove the cut points selected by our study.

Conflict of interest

The authors have no conflict of interest to declare.