Effect of peri-operative blood transfusions on long-term prognosis of patients with colorectal cancer

Abstract

Background

Patients with colorectal cancer often present with anaemia and require red blood cell transfusions (RBCT) during their peri-operative course. Evidence suggests a significant association between RBCT and poor long-term outcomes in surgical patients, but the findings in colorectal cancer are contradictory.

Material and methods

The aim of this retrospective, single-centre, cohort study was to investigate the prognostic role of peri-operative RBCT in a large cohort of patients with stage I–III colorectal cancer submitted to curative surgery between 2005 and 2017. The propensity score matching technique was applied to adjust for potential confounding factors.

Results

Among 1,414 patients operated within the study period, 895 fulfilled the inclusion criteria: 29.6% (n=265) received peri-operative RBCT. The group that received peri-operative RBCT was significantly older (p<0.001), had more comorbidities (p<0.001), more advanced tumours (p<0.001) and more colon tumours (p=0.002) and stayed in hospital longer (p<0.001). Post-operative mortality was 7-fold higher (2.3 vs 0.3%, p=0.01) in this group. Survival outcomes were significantly worse in the group receiving RBCT than in the group not receiving RBCT for both overall (64.5 vs 80.1%, p<0.001) and cancer-specific survival (74.3 vs 85.1%, p<0.001). On multivariable analysis, peri-operative RBCT was significantly associated with poorer overall survival (hazard ratio 1.51, p=0.009). When transfused and non-transfused cases were paired through the propensity score matching technique considering main clinico-pathological features, no differences in overall and cancer-specific survival were found.

Discussion

Our data suggest that, after adjustment for potential confounding factors, no significant association exists between RBCT and prognosis in colorectal cancer.

INTRODUCTION

Colorectal cancer (CRC), one the most common tumours worldwide¹, frequently leads to multifactorial anaemia^2,3, which is in part caused by bleeding from the tumour⁴ and by release of inflammatory cytokines⁵. Occult or overt gastrointestinal blood loss, bone marrow depletion, iron deficiency, and side effects of neoadjuvant therapy are among other factors that can be responsible for anaemia in this population of patients⁶. Up to 60% of patients with CRC currently require at least one red blood cell (RBC) transfusion during their surgical course⁷. In addition to the economic burden^8,9, evidence suggests a significant association between blood transfusion and poor short- and long-term outcomes in surgical patients^10–13. Blood transfusions have been associated with poorer prognosis in patients with lung¹⁴, head and neck¹⁵, and ovarian cancers¹⁶. Three recent meta-analyses also showed a moderate association between peri-operative autologous blood transfusion and increased risk of tumour recurrence in patients with CRC¹⁷, along with adverse clinical outcomes, including increased mortality^18,19. Nevertheless, the studies included in these meta-analyses were mostly heterogeneous, since large variability in the timing of transfusions and definitions of transfused blood components were commonplace, and the sample size was generally limited. Moreover, adjustment for relevant confounding factors such as comorbidities, disease stage or other pathological features was often overlooked. The correlation between blood transfusion and worse long-term outcomes seems therefore due to an indirect and non-causal relationship.

The primary purpose of this single-institution, retrospective study was to evaluate the prognostic role of peri-operative RBC transfusion (PeriBT) in a large cohort of patients with stage I–III CRC submitted to curative surgery. The clinical and pathological characteristics of patients receiving RBC transfusions were also investigated.

MATERIALS AND METHODS

Inclusion criteria and population under study

The study population consisted of all patients undergoing surgery for colorectal tumour (1,414 cases) at the Division of General and Hepatobiliary Surgery, University of Verona Hospital, between January 2005 and December 2017. Inclusion criteria were age 18 years or older, histology-proven colorectal tumour, absence of metastasis, potentially curative resection, elective surgery, availability of transfusion history, and minimum follow-up of 12 months. Patients undergoing palliative surgery or with missing pathological data were excluded. A CONSORT diagram showing the process of patients’ selection and inclusion into the study is provided in Figure 1. A total of 895 patients fulfilled the inclusion criteria and entered the study. Patients were thereafter classified into two groups, according to their transfusion status (PeriBT vs non-PeriBT). A preliminary analysis showed relevant differences between the two groups in terms of demographic and clinical characteristics. We therefore decided to use the propensity score matching technique to adjust for potential confounding factors, obtaining two homogeneous groups of 165 patients each for the most relevant demographic characteristics, clinical features, and surgical data. All methods used in this study were performed in accordance with the ethical guidelines and regulations of the University Hospital of Verona, in which the investigation was carried out. Informed consent was obtained from all patients and the study protocol was approved by the local Ethical Committee (ID number: 42763 - CRINF-1034 CESC).

Figure 1.

CONSORT diagram showing the process of selecting and matching patients

RBC: red blood cells.

Pre-operative work-up and histopathological staging

Prior to surgery, all patients were staged with colonoscopy, chest-abdomen-pelvis computed tomography and measurement of carcinoembryonic antigen. Additional imaging studies, including magnetic resonance and transrectal ultrasound, were used for staging rectal cancer. Liver magnetic resonance and positron emission tomography were also used to evaluate uncertain lesions in stage IV cases. Pathology specimens were reported according to the 7^th Edition of the American Joint Committee on Cancer and the Union International Contre Le Cancer criteria.

Extent of surgery

Complete excision of the tumour burden (R0 resection) was the main goal of surgery. The extent of surgery was planned according to each patient’s condition, tumour location and stage. Both in open and laparoscopic surgery, anatomical CRC resections with ligation of vessels at their origin were preferred in order to harvest an adequate number of lymph nodes.

Assessment of haemoglobin levels and transfusion history

Pre-operative haemoglobin values were determined within 2 weeks prior to the date of surgery and repeated the day before surgery in patients with severe anaemia. Data regarding haemoglobin levels were obtained by querying the Medical Laboratory database system into which all laboratory results were stored. Pre-operative anaemia (suboptimal haemoglobin level) was defined as a haemoglobin concentration of less than 130 g/L for males and less than 120 g/L for females, as defined by World Health Organization (WHO)²⁰. According to haemoglobin levels on presentation, pre-operative anaemia was subclassified as mild (haemoglobin 110–129 g/L in males and 110–119 g/L in females) or moderate (<10⁹ g/L).

PeriBT was defined as a transfusion of RBC from 30 days before (POD -30) to 90 days (POD 90) after surgery, and data were obtained by reviewing the Blood Transfusion Service database. The need for transfusions was determined on a case by case basis, according to clinical course and haemoglobin levels. Generally, a haemoglobin of 80 g/L was considered as a cut-off value in low-risk patients, with values below indicating the need for a transfusion, while a value <100 g/L was considered for high-risk patients.

Data collection and statistical analysis

All clinical and pathological data from a prospectively compiled digital database were retrospectively analysed. Demographic, clinical, surgical and pathology variables were analysed. Comorbidity status was assessed using the American Society of Anesthesiologist (ASA) Physical Status Classification System²¹ and the Charlson-Age Comorbidity Index (CACI)²². Type and severity of peri-operative complications were recorded within 30 days of primary surgery, and classified according to the Clavien-Dindo classification²³. Post-operative mortality was defined as death occurring within 30 days after the operation or during the hospital stay. Survival and follow-up data were obtained by collecting outpatient clinical records, or by directly contacting the patient. The minimum follow-up period was 24 months. Transfused and non-transfused cases were paired using the propensity score matching technique with the aim of obtaining two homogeneous groups for year of surgery, age, gender, tumour location, CACI score, pre-operative haemoglobin value, tumour-node-metastasis (TNM) stage and post-operative complications (Online Supplementary Figure S1).

Continuous data were reported as mean (+ standard deviation) or median (range) as appropriate according to distribution, whilst categorical data were reported as numbers and percentages. Comparisons between groups were carried out with an independent t test or Mann-Whitney U test for continuous variables and the chi-squared test or Fisher’s exact test for categorical variables, as appropriate. All statistical tests were two-sided, and association were considered statistically significant at a nominal level of 0.05 (p<0.05). Survival analysis was computed using the Kaplan-Meier method and compared by the log-rank test with time of overall survival (OS), measured from the date of surgery to the date of death from any cause or most recent follow-up. Cancer-specific survival (CSS) was measured from the date of surgery to the date of death from cancer. Multivariable analysis for OS and CSS was performed using Cox regression model considering transfusion status (PeriBT vs no transfusion) and adjusting for the following risk factors: age (≤ median vs > median), gender (male vs female), CACI score (≤ 4 vs >4), pre-operative anaemia (no anaemia vs anaemia), histopathologically determined tumour status (pT1–pT2 vs pT3 vs pT4), and histopathological determined lymph node status (pN0 vs pN1 vs pN2). Statistical analysis was performed using SPSS, version 23 (SPSS, Chicago, IL, USA).

RESULTS

From an initial population of 1,414 patients assessed for eligibility, 895 fulfilled the inclusion criteria and were hence analysed (Figure 1). Overall, 265 (29.6%) patients received at least one RBC transfusion before (n=119), during (n=85), or after (n=196) surgery. The median number (interquartile range [IQR]) of transfused units was two (IQR, 2–4), with a maximum of 20 units in one patient. Table I reports the clinical, pathological and surgical characteristics of the population. Patients in the PeriBT group were significantly older (p<0.001), had higher CACI score (p<0.001), and were more frequently classified as ASA 3 or above (p<0.001). As expected, the median pre-operative haemoglobin concentration was lower in patients who needed RBC transfusions (109 g/L vs 136 g/L, p<0.001), with 79.8% of these patients being anaemic. Regarding pathological characteristics, tumours in the PeriBT group were mainly located in the colon (p=0.002), more frequently invading the pericolic tissues and the serosa (p<0.001) and presented at more advanced stages (p<0.001). Finally, patients who receivedperi-operative RBC transfusions stayed in hospital longer after surgery (p<0.001) and more often developed post-operative complications (p<0.001), including those graded as Clavien-Dindo III or higher (p<0.001). Post-operative mortality was 7-fold higher in patients who received PeriBT (2.3 vs 0.3%, p=0.01).

Table I.

Demographic and clinical-pathological characteristics for the 895 patients under study

Data	No periBT (n=632)	PeriBT (n=263)	p value
Age, years, median (IQR)	66.5 (58.3–74.7)	75.1 (64.1–81.6)	<0.001
Gender, female, n (%)	260 (41.1)	118 (44.9)	0.334
Tumour location, rectum, n (%)	239 (37.8)	71 (27.0)	0.002
BMI, kg/m 2 , median (IQR)	25.5 (23.3–28.0)	24.9 (22.5–27.2)	0.097
ASA physical status class ≥3, n (%)	139 (22.0)	136 (52.4)	<0.001
CACI score, median (IQR)	4 (3–5)	5 (4–6)	<0.001
Pre-operative Hb, g/L, median (IQR)	136 (123–148)	109 (97–121)	<0.001
Laparoscopic surgery, n (%)	198 (31.3)	48 (18.3)	<0.001
N. of excised nodes, median (IQR)	19 (13–26)	19 (14–28)	0.435
Depth of tumour invasion (pT), n (%)			<0.001
pT1–T2	237 (37.5)	56 (21.3)
pT3	282 (44.6)	129 (49.0)
pT4	113 (17.9)	78 (29.7)
Nodal involvement (pN), n (%)			0.137
pN0	425 (67.2)	159 (60.5)
pN1	152 (24.1)	74 (28.1)
pN2	55 (8.7)	30 (11.4)
AJCC TNM stage, n (%)			<0.001
Stage I	190 (30.1)	39 (14.8)
Stage II	235 (37.2)	120 (45.6)
Stage III	207 (32.8)	104 (39.5)
30-day post-operative morbidity, n (%)	112 (24.2)	92 (46.0)	<0.001
30-day post-operative major morbidity, n (%)	13 (2.8)	37 (18.6)	<0.001
30-day post-operative mortality, n (%)	2 (0.3)	6 (2.3)	0.01
Length of hospital stay, days, median (IQR)	8 (7–10)	11 (8–15)	<0.001

PeriBT: peri-operative red blood cell transfusion; IQR: interquartile range; BMI: body mass index; ASA: American Society of Anesthesiologists; CACI: Charlson-Age Comorbidity Index; Hb: haemoglobin; AJCC: American Joint Committee on Cancer; TNM: tumour, node, metastasis.

PeriBT was significantly associated with poorer 5-year OS (64.5 vs 80.1%, p<0.001) (Figure 2A) and CSS (74.3 vs 85.1%, p<0.001) (Figure 2B). No differences were observed in recurrence rate (20.7 vs 16.8%, p=0.18). Multivariable analysis for OS and CSS was initially conducted in the whole population considering transfusion status and adjusting for several risk factors (Table II). In this model, age above the median and TNM stage were confirmed to be significant predictors of OS and CSS. The administration of PeriBT was significantly associated with poorer OS (hazard ratio [HR] 1.51; 95% confidence interval [95% CI]: 1.11–2.04, p=0.009), but did not influence CSS.

Figure 2.

Kaplan-Meier estimates of overall and cancer-specific survival

(A) Overall survival in the unmatched population. (B) Cancer-specific survival in the unmatched population. (C) Overall survival in the matched population. (D) Cancer-specific survival in the matched population.

Table II.

Multivariable analysis for overall survival and cancer-specific survival in the unmatched population

	Overall survival		Cancer-specific survival
	OR (95% CI)	p	OR (95% CI)	p
Age		0.002		0.04
≤median (68.5 years)	1		1
>median	2.09 (1.32–3.32)		1.76 (1.03–2.99)
Gender		0.16		0.24
Male	1		1
Female	0.83 (0.64–1.07)		0.84 (0.62–1.15)
CACI score		0.01
≤4	1		1
>4	1.87 (1.15–3.03)		1.68 (0.96–2.95)
Pre-operative anaemia		0.44		0.84
None	1		1
Mild-moderate	1.13 (0.83–1.53)		1.04 (0.72–1.51)
TNM stage		<0.001		<0.001
Stage 1	1		1
Stage 2	1.41 (0.95–2.07)		1–70 (1.02–2.83)
Stage 3	2.37 (1.11–3.45)		3.36 (2.06–5.50)
Peri-operative RBC transfusion		0.009		0.08
No	1		1
Yes	1.51 (1.11–2.04)		1.40 (0.96–2.03)

OR: odds ratio; 95% CI: 95% confidence interval; CACI: CACI: Charlson-Age Comorbidity Index; TNM: tumour, node, metastasis; RBC: red blood cell

After application of the propensity score technique, 165 cases in the PeriBT group were matched to 165 controls. Online Supplementary Figure S1 shows the propensity score graphs of the unmatched and matched groups. In the propensity score matched cohort (Online Supplementary Table SI), the demographic, pathological and post-operative characteristics of the patients in the PeriBT group were similar to those of the control group. The only differences were lower body mass index (24.2 vs 25.8 kg/m², p=0.007) and longer stay in hospital(10 vs 8 days, p<0.001). There was no significant difference in 5-year OS (67.3 vs 67.9%, p=0.28) or CSS (77.9 vs 77.1%, p=0.59) after matching (Figure 2C, D). Similarly, multivariable analysis did not show any association between PeriBT and OS or CSS (Table III).

Table III.

Multivariable analysis for overall survival and cancer-specific survival after matching

	Overall survival		Cancer-specific survival
	OR (95% CI)	p	OR (95% CI)	p
Age		0.42		0.66
≤median (68.5 years)	1		1
>median	1.37 (0.63–2.96)		1.24 (0.47–3.22)
Gender		0.19		0.31
Male	1		1
Female	0.78 (0.53–1.14)		0.78 (0.48–1.26)
CACI score		0.02		0.21
≤4	1		1
>4	2.68 (1.16–6.22)		1.90 (0.69–5.24)
Pre-operative anaemia		0.45		0.28
None	1		1
Mild-moderate	0.86 (0.58–1.28)		0.76 (0.47–1.24)
TNM stage		0.02		0.005
Stage 1	1		1
Stage 2	1.03 (0.60–1.78)		1.72 (0.76–3.92)
Stage 3	1.73 (1.01–2.30)		3.20 (1.41–7.25)
Peri-operative RBC transfusion		0.34		0.64
No	1		1
Yes	1.20 (0.83–1.74)		1.12 (0.70–1.78)

OR: odds ratio; 95% CI: 95% confidence interval; CACI: CACI: Charlson-Age Comorbidity Index; TNM: tumour, node, metastasis; RBC: red blood cell.

DISCUSSION

The main results of this study are: (i) PeriBT were more frequently administered to elderly patients, with comorbidities, locally advanced lesions located in the colon and consequent pre-operative anaemia; (ii) patients receiving PeriBT had higher rates of post-operative complications and mortality, along with longer stays in hospital; (iii) PeriBT were associated with poorer OS and CSS in the unmatched cohort of patients undergoing curative surgery for stage I–III CRC; (iv) after matching for clinical and pathological characteristics, PeriBT did not have a significant influence on either OS or CSS.

The findings of the retrospective studies published so far on the influence of PeriBT on the prognosis of patients with CRC are contradictory, being nearly equally divided between those reporting that transfusions may be a significant factor^{12,17,24–28} and those stating that transfusions do not represent a statistically significant prognostic indicator^9,29–35. It is still unclear whether the worsening of long-term outcomes may be consequent to the transfusion itself, or could be attributed to causal circumstances, more specifically to pre-operative anaemia¹², higher tumour stage or increased surgical difficulty. Moreover, there is great variability in the characteristics of patients included in the studies and, even more importantly, the definition of peri-operative blood transfusion is still heterogeneous and poorly standardised²⁷. The lack of consistency in results from the various studies might also be attributable to the different study designs and haemoglobin levels used to define pre-operative anaemia, as well as complex treatment-related factors such as the surgical approach, the amount of intra-operative bleeding, and the indications for blood transfusion³⁶. Finally, current evidence suggests that duration and type of anaesthesia³⁷, amount of blood loss, post-operative complications³⁸, and systemic inflammation in the peri-operative period^39–41 probably all play important roles in the fate of residual disease. Interestingly, patients in the PeriBT group before matching had all of these risk factors; specifically, they had more advanced disease stages, pre-operative anaemia, and a higher rate of post-operative complications. For these reasons, we decided to match the patients who received PeriBT with those who did not but who had similar clinical and pathological characteristics. The aim of matching was to mitigate the influence of potential confounding factors over the effect of blood transfusions. Indeed, no significant association was found between PeriBT and OS or CRS after correcting for year of surgery, age, gender, tumour location, preoperative Hb value, comorbidity status, TNM stage and post-operative complications.

Although blood transfusion is lifesaving in many circumstances for improving the oxygen-carrying capacity of blood, it also entails non-negligible risks, including incompatibility, transmission of infectious agents and allergic reactions⁴². It is also well-known from transplantation literature that blood transfusion improves the survival of renal allografts, with a dose-response relationship⁴³, thus suggesting an immunosuppressive effect of allogenic blood transfusion. After the discovery of this lasting immunosuppressive effect, concern has been expressed about the potential negative consequences on cancer patients^24,44. The most popular hypothesis explaining the dose-response relationship between blood transfusions and improved survival of renal allografts is that the leucocytes in transfused blood stimulate the production of non-specific suppressor cells in the recipient, which would in turn attenuate the cellular immune reaction against the graft^45–48. While this immunosuppression seems beneficial in organ transplantation, it may be harmful in oncological surgery, in which immune surveillance is important for detecting and eliminating circulating malignant cells²⁷. Inflammatory and immunosuppressive mediators were found to be associated with recurrences and metastasis, while transfusion could also accelerate tumour progression by inducing an inflammatory response and immunosuppression¹⁹. Interestingly, the first studies evaluating the effect of blood transfusion on cancer patients considered blood components that are now rarely used, such as whole blood and packed RBC^25–27,32. Packed RBC contain about 50–60 mL of plasma and anticoagulant preservative solution, whole blood contains 250–300 mL, or five to six times more. RBC units may also be partially depleted of lymphocytes, but leucodepleted RBC were not commonly used in the past²⁵. In our cohort, the vast majority of patients received leucodepleted RBC units, and none was transfused using whole blood. The results of the oldest studies are thus not fully comparable with ours, since the difference in outcomes may be explained by the detrimental effect of the leucocytes present in whole blood^12,48.

Anaemia per se could also negatively impact the prognosis of malignancies, and could increase the risk of overall mortality⁴⁹. The presence of pre-operative anaemia in CRC surgery has been identified as an independent risk factor for post-operative complications, longer hospital stay⁶ and poorer long-term prognosis^12,50,51. In our experience, pre-operative haemoglobin level was significantly lower in the PeriBT group before matching, and this difference may partly contribute to the difference in survival in the unmatched population.

Our study has some limitations, mainly related to its retrospective nature and some missing data. The lack of a formal blood transfusion protocol during the study period means that the decision to transfuse was made by the surgeons or anaesthetists. On the other hand, our study has also several strengths. First, the cohort is one of the largest in similar clinical studies. Second, although propensity score matching significantly reduced the number of patients, it allowed us to control for relevant confounding factors, including pre-operative haemoglobin level, year of surgery, tumour features and patients’ characteristics. Since a randomised controlled trial based on the administration or withholding of blood transfusion would be unethical, the use of propensity score matching to adjust for potential risk factors becomes of primary interest.

CONCLUSIONS

The results of the present study confirm that RBC transfusion is frequent in patients undergoing surgery for CRC, mostly in those with advanced age, pre-operative anaemia and several comorbidities. However, after adjustment for relevant confounding factors, no significant causal association was found between PeriBT and OS and CSS in patients undergoing curative surgery for stage I–III CRC.