Preoperative anaemia and perioperative red blood cell transfusion as prognostic factors for recurrence and mortality in colorectal cancer—a Swedish cohort study

Preoperative anaemia, present in approximately 50 % of the CRC patients in this study, was a strong predictor for impaired survival. There was no significantly increased risk of cancer recurrence in patients with preoperative anaemia. Blood transfusion was neither associated with increased risk for recurrence, nor with overall mortality in the present study.

This is a single centre retrospective cohort study using prospectively registered registry data. Consecutive patients undergoing radical abdominal resection for stages I–III CRC at the Karolinska University Hospital and living in the region were included. The minimum follow-up time in the study was 3 years. Karolinska University Hospital is one of the largest centres in Sweden, taking advanced cases from the Stockholm Region. Patients referred from other regions were not included in the trial since they are followed up at their nearest hospital.

For each patient, the lowest of all haemoglobin values registered within 2 months before the date of surgery was chosen for classification of anaemia. This was to minimise the risk of patients receiving preoperative iron medication or blood transfusion and being misclassified.

Data on blood transfusion was collected 2 months before surgery until 1 month after surgery. Analyses were performed for patients receiving blood transfusions within 1 day before or after surgery, since this time interval was considered the most interesting period with respect to impaired immunological response of the patient at the time of surgical trauma. One hundred ninety-nine patients (81 % of patients receiving blood transfusion) did so within this time interval. It can be argued that the immunological effect of blood transfusion in combination with surgical stress can be longer. Additional analyses for the period 1 month before or after surgery were performed regarding recurrent disease and survival.

The only available measure of co-morbidity was ASA class. This factor was adjusted for in the multivariate models since it was suspected to influence both survival and the risk for anaemia.

Abdomino-perineal resection is known to be associated with a higher risk for recurrence [21], hence the analyses were adjusted for surgical procedure. Although, such an association could not be shown in the current study. Analyses were also performed when dividing colonic resections into acute and elective procedures [22]. This did not affect the outcome estimates, and colon resections were thus analysed as one group.

The stage reported to the SCRCR is the stage derived from examination of the pathology specimen. Since 80 % of the patients with rectal cancer received neo-adjuvant (chemo) radiotherapy, and in total 36 % of all patients received any neo-adjuvant oncological treatment, down staging might have occurred in about a third of patients.

The risk for recurrence was associated with preoperative anaemia in the log-rank test but the confidence interval in the multivariate analysis did just include 1 (0.97–2.6) and was thus not significant. However, anaemia was significantly associated with the risk for recurrence in the model including both anaemia and transfusion. Therefore, this could be described as a trend and possibly this would have been significant in both models if the sample size was larger. More advanced T and N stage was associated with an increased risk for recurrence in the multivariate model. The association between anaemia and risk for recurrence has previously been shown in a recent study on stage II CRCs where anaemia was shown to be associated with a reduction in recurrence-free survival in T3N0M0 colon cancer patients only, but not in patients with T4N0M0 colon cancers [13]. In that study, adjustment was made for tumour but not for patient characteristics.

Blood transfusion, often defined in studies as given within 30 days before and after surgery, has repeatedly been found to be associated with an increased risk for recurrence [10]. Although there is little specific data on the mechanisms behind such an association, it has typically been attributed to the immunological side effects of blood transfusion [23]. The recurrence risk described previously was independent of tumour stage or when the transfusion was administered [10]. In the present study, no association was found between transfusion and risk for recurrence. The effect of transfusions administered within a day or 30 days before or after surgery showed practically identical results.

Anaemia prior to surgery, gender, ASA and pN-stage class were significantly associated with increased overall mortality regardless of tumour stage. The present study only included CRC patients stages I–III with curative abdominal resection, where all early postoperative deaths were excluded, making the cohort homogenous. Anaemia was associated with increased overall mortality in both the analysis for anaemia alone and in the combined analysis for anaemia and transfusion. The multivariate analyses revealed no association between transfusion and increased overall mortality although the log-rank test of the univariate Kaplan-Meyer model was significant.

Little is known about preoperative anaemia and the risk for death in patients operated for CRC [11‐13]. Previous studies have shown an association between mild anaemia preoperatively and reduced overall survival 3–8 years after surgery in patients with CRC stages I–IV [12]. That study focused on the association between preoperative symptoms and survival [12]. Mild anaemia was associated with a more advanced stage but stage was not entered into the multivariate analyses [12]. The negative effect of preoperative anaemia on survival was also confirmed in another study including patients after both curative and palliative surgery [11].

Hazard ratios were virtually unaffected when analysing anaemia and transfusion as independent variables in the various multivariate models compared to HRs in the combined model although the association between anaemia and recurrence did not reach statistical significance in the model including only anaemia. Preoperative anaemia was, in both models, a significant risk factor for overall mortality. We were unable to show that transfusion has an impact on these outcomes in the multivariate models. It can be argued not to include both anaemia and blood transfusion in the same model since they are dependent of each other but, we believe, that showing both the separate models and the combined model increases the robustness of the results. The results of the present study stand in stark contrast to the conclusions of two meta-analyses where perioperative blood transfusion was considered to be consistently associated with risk for recurrence [10, 24]. One possible explanation could be that the Karolinska University hospital began using leucocyte-depleted blood components in 1997, so that all patients included in this study received these presumably less immune-modulatory blood components [23]. In addition, one must also consider the possibility of our study being under powered. While this may be indicated by the comparatively low statistical precision it must be kept in mind that the upper confidence interval for the estimate of the effect of blood transfusion on the risk for recurrence was 1.3. This would imply a lesser effect than that derived in the previous meta-analyses [10, 23].