Samuel D. Kaffenberger, Department of Urologic Surgery, Vanderbilt University Medical Center, A1302 Medical Center North, Nashville, TN 37215, USA. e-mail: email@example.com
Study Type – Prognosis (cohort series)
Level of Evidence 3a
What's known on the subject? and What does the study add?
Some evidence suggests that ABO blood type may be a risk factor for cancer incidence and prognosis. For example, a large study recently discovered an increased incidence of pancreatic cancer in patients with non-O blood type; however, it is not known whether blood group correlates with outcomes in patients with RCC.
We found a significant and independent association between ABO blood group and overall survival in patients undergoing surgery for locoregional RCC. Specifically, we identified non-O blood type as a predictor of mortality.
• To determine whether ABO blood group is associated with survival after nephrectomy or partial nephrectomy for renal cell carcinoma (RCC).
PATIENTS AND METHODS
• We conducted a retrospective cohort study of 900 patients who underwent surgery for locoregional RCC between 1997 and 2008 at a single institution.
• Covariates included age, gender, race, American Society of Anesthesiology Physical Status, preoperative anaemia and hypoalbuminemia, tumour characteristics, lymph node status, procedure performed, transfusion status and ABO blood group.
• Primary outcomes were overall (OS) and disease-specific survival (DSS).
• Univariable survival analyses were performed using the Kaplan–Meier and log-rank methods. Multivariable analysis was performed using a Cox proportional hazards model.
• The 3-year OS estimate was 75% (95%CI 70–79%) for O blood group and 68% (95% CI 63–73%) for non-O blood group (P= 0.072). The 3-year DSS was 81% (95% CI 76–85%) for O blood group and 76% (95%CI 71–80%) for non-O blood group (P= 0.053).
• In the multivariable analysis for OS, non-O blood type was significantly associated with decreased OS (HR 1.68, 95%CI 1.18–2.39; P= 0.004) but not DSS (HR 1.53, 95%CI 0.97–2.41; P= 0.065).
• These data suggest that ABO blood group is independently associated with OS in patients undergoing surgery for locoregional RCC. ABO blood group has not been previously recognized as a predictor of survival in RCC.
American Society of Anesthesiology Physical Status
Kidney cancer is expected to account for >13 000 deaths in the USA in 2011 and resulted in ∼ 25 000 radical and 10 000 partial nephrectomies in 2009 . A number of factors that predict survival in patients with locoregional RCC after radical or partial nephrectomy have been elucidated, including age, race, gender, performance status, stage, nutritional status and tumour size [2–4]. Recent work in genetics and molecular biomarkers has also been promising . Although multiple genetic syndromes influence both the incidence and survival of patients with RCC, the majority of prognostic factors are non-heritable . Despite these advances, clinically useful biomarkers or heritable risk factors for RCC prognosis outside of familial syndromes have remained elusive.
The possibility that ABO blood group status may be a cancer risk factor was initially proposed over 50 years ago, and there has been a renewed interest in this question [7,8]. A number of early, mostly small, retrospective case–control studies explored the relationship between blood group and gastric, uterine, bladder and other cancers [7,9,10]. More recently, non-O blood type was associated with an increased risk of pancreatic cancer in two large independent populations ; this study suggested that 17% of pancreatic cancers were directly attributable to non-O blood type and the findings were subsequently validated in a separate meta-analysis . While most of these studies found no relationship between blood group and survival, a contemporary study found O blood type to be associated with improved survival compared with non-O blood type in patients with breast cancer after mastectomy . The few studies that have examined ABO blood group and incidence of RCC have shown conflicting results, and all suffer from substantial methodological shortcomings [13,14]; therefore, whether ABO blood group is associated with RCC incidence has not been properly explored and its relationship with prognosis has never been studied.
Any association between blood group and survival in patients undergoing surgery for RCC would have important implications. While not necessarily guiding treatment decisions, recognition of a prognostic, heritable trait that is routinely assessed in these individuals could affect clinical counselling and point to new avenues of molecular research for understanding the biology of RCC progression. We therefore sought to study the association between ABO blood group and survival in patients with locoregional RCC undergoing radical or partial nephrectomy. To our knowledge, this represents the first study to assess the usefulness of ABO blood group as a predictor of outcome in RCC. We hypothesized that O blood type would be associated with improved survival in these patients.
PATIENTS AND METHODS
We performed a retrospective cohort study of 900 consecutive patients who underwent radical or partial nephrectomy for locoregional RCC between 1997 and 2008 with recorded ABO blood type. All RCC histological subtypes were included. Patients with locoregional advanced disease such as direct adrenal invasion, regional lymph node metastases, or tumour thrombus were included; however those with distant metastases were excluded from the analysis. All surgeries were performed at Vanderbilt University Medical Center, with preoperative evaluation and postoperative follow-up performed according to institutional protocol. Cause of death was determined by the treating physicians, death certificate and/or chart review.
A staff surgical pathologist evaluated all surgical specimens. Stage and grade were assigned according to the 2010 American Joint Committee on Cancer guidelines and Fuhrman grading system, respectively. Clinical, pathological and outcome data were collected prospectively and were supplemented by medical record review. Institutional review board approval was obtained for the creation of a prospective database and for retrospective analysis of this patient population.
The primary outcome measures in this study were overall (OS) and disease-specific survival (DSS). Duration of follow-up was the time from surgery to the date of death or last clinic visit.
We evaluated clinical and pathological variables including age, gender, race (white vs. non-white), American Society of Anesthesiology Physical Status (ASA-PS) score, preoperative anaemia (haematocrit <41 for men and <36 for women), preoperative hypoalbuminemia (albumin <3.5 g/dL), Fuhrman nuclear grade (I–II vs. III–IV), pathological T-stage, node status, tumour histology (clear-cell vs. non-clear-cell), procedure performed (radical vs. partial nephrectomy), red blood cell transfusion status, and ABO blood group. Patients without radiographic or palpable evidence of lymphadenopathy generally did not undergo lymphadenectomy(Nx) and were grouped with pathological N0 patients for analysis.
The relationship between ABO blood group and clinicopathological variables was assessed using chi-squared tests. Patients were stratified for analysis as either type O or type non-O, based upon the study showing a link between non-O blood type and decreased survival in patients with breast cancer after mastectomy . Univariable survival analyses were performed using the Kaplan–Meier and log-rank methods. Cox proportional hazards models for OS and DSS were constructed for the multivariable survival analyses. Based on previous analysis of this institutional dataset, covariates were age, race, ASA-PS score, pT stage, Fuhrman grade, node status, preoperative anaemia, and preoperative hypoalbuminemia . Charlson comorbidity index score was only available for the most recent subset of patients, therefore ASA-PS score was used as an indicator of comorbidity. A total of 567 patients had complete information for all variables, and these patients were included in the multivariable survival analyses. All analyses were conducted with stata data analysis software (College Station, TX, USA, version 11).
The median (interquartile range [IQR]) age of the cohort was 61 (51–69) years. The median (IQR) follow-up was 28.7 (15.1–49.8) months. The median (IQR) follow-up of patients alive at last follow-up was 34.1 (19.0–54.6) months. Table 1 gives the distribution of clinicopathological variables by ABO status. Blood group is randomly distributed throughout all clinicopathological variables with the exception of race, where non-white individuals had a higher prevalence of type O blood (P= 0.021). Table 2 gives the distribution and 3-year OS and DSS probabilities according to blood type.
Table 1. Distribution of patients by clinical and pathological variables according to blood group
All P values from chi-squared test. Percentages may not add up to 100% owing to missing data.
Table 2. Distribution of patients and 3-year survival probabilities by blood group
3-year OS (95%CI), %
3-year DSS (95%CI), %
The actuarial 3-year OS for our entire cohort of 900 patients was 73.0% (95% CI 69.6–76.0% [Table 2]). Of the 567 patients with complete data included in the multivariable analysis, there were 137 overall deaths and 83 disease-specific deaths. The 3-year OS estimate was 75% (95%CI 70–79%) for O blood group and 68% (95%CI 63–73%) for non-O blood group (P= 0.072). The 3-year DSS was 81% (95%CI 76–85%) for O blood group and 76% (95%CI 71–80%) for non-O blood group (P= 0.053). The Kaplan–Meier survival analyses for OS and DSS by blood type are shown in Figs 1 and 2, respectively.
In the univariable analysis, age, ASA-PS score, pT stage, node positivity, grade, anaemia and hypoalbuminemia were all significantly associated with OS (Table 3). Blood group was not significant in the univariable analysis for OS (hazard ratio [HR] 1.24 for non-O vs. O, 95% CI 0.98–1.58; P= 0.072 [Table 3]). In the univariable analysis for DSS, higher stage and grade, node positivity, anaemia and hypoalbuminemia were all significantly associated with decreased DSS (Table 4). Non-O blood group compared with O blood group approached significance for decreased DSS (HR 1.34, 95%CI 0.99–1.82; P= 0.054 [Table 4]).
Table 3. Univariable and multivariable Cox proportional hazards regression for OS
Blood group (non-O vs O)
Table 4. Univariable and multivariable Cox proportional hazards regression for DSS
Blood group (non-O vs O)
In the multivariable analysis for OS, higher stage and grade, higher ASA-PS class, node positivity, anaemia and hypoalbuminemia, were all significantly associated with OS (Table 3). In addition, non-O blood group was significantly associated with an increased risk of overall mortality (HR 1.68, 95%CI 1.18–2.39; P= 0.004 [Table 3]). In the multivariable analysis for DSS, stage, grade, node positivity, anaemia and hypoalbuminemia were significant predictors of mortality (Table 4).The relationship between blood group and DSS was not significant (HR 1.53, 95%CI 0.97–2.41; P= 0.065 [Table 4]).
As an exploratory analysis, the multivariable model for OS was assessed with each blood type entered separately rather than grouped as O vs. non-O. Relative to the O blood type, both A (HR 1.60, 95%CI 1.10–2.34, P= 0.014) and AB (HR 2.89, 95%CI 1.13–7.39, P= 0.027) blood types were independently associated with OS while B blood type (HR 1.60, 95%CI 0.88–2.92, P= 0.127) was not a significant predictor of OS on its own. Similarly, in an exploratory multivariable model for DSS, only AB blood type was associated with DSS (HR 3.37, 95%CI 1.14–9.96, P= 0.028).
Our data suggest that ABO blood group is independently associated with OS in patients undergoing surgery for locoregional RCC. Specifically, we identified non-O blood type as a potential predictor of mortality in these patients, as patients with non-O blood types had a 68% increased mortality risk compared with those with blood type O. In both the univariable and multivariable analyses for DSS, non-O blood type approached significance (P= 0.054 and P= 0.062, respectively). To our knowledge, this is the first reported association between ABO blood group and OS in RCC.
An association between the incidence of gastric cancer and ABO blood groups was described as early as 1953 . Since that time, studies on ABO blood groups and cancer have had conflicting results, though few have identified a link between blood group and survival. In a series of 315 women with breast cancer, Costantini et al. showed that ABO blood group was significantly associated with OS and DSS after mastectomy . Consistent with the present findings, women with non-O blood type had decreased survival after radical surgery.
No previous studies have examined whether there is an association between blood type and prognosis in RCC and those that have evaluated RCC incidence and blood type have often been limited by cohort size and methodological approach. A retrospective case–control study from 1976 compared the blood groups of 125 patients with RCC who underwent surgery over a 30-year period with the normal blood group make-up of Northeastern Pennsylvania. They found an increase in incidence of RCC in patients with A-, O-, and B+ blood types and a decrease in incidence in patients with O+ blood . However, conclusions are difficult to draw from this study given the lack of rigorous statistical analysis, small numbers and lack of racial and socio-demographic information. Another small retrospective case–control study from 1985 evaluated 93 patients with RCC and did not show any significant effect of blood type on RCC incidence . A recent analysis of patients with RCC undergoing treatment at the European Institute of Oncology did not show any statistically significant difference in distribution of blood types for patients with RCC compared with controls .
Some mechanisms have been previously proposed to account for the possible effects of ABO blood type on cancer incidence and prognosis, and our findings are hypothesis-generating for potentially novel RCC translational research avenues. The ABO blood group alleles are found on chromosome 9q34.1 and encode specific glycosyltransferases which alter a protein backbone, the H antigen, to create the A and B cell surface antigens . Owing to a frameshift mutation, the O allele encodes a non-functional glycosyltransferase and the H antigen is thereby unaltered in this instance. In addition to their expression on the surface of red blood cells, the ABO antigens are expressed on the cell surface of a number of epithelial cell types, including the urogenital epithelium . Notably, these ABO cell surface antigens have been shown to be differentially deleted or altered in neoplastic urothelium . Furthermore, aberrant glycosylation is considered to be a hallmark of human cancer and while RCCs do not appear to express the ABO cell surface antigens, alterations in tumour expression of the A and B glycosyltransferases could affect surrounding epithelial and endothelial cell surface antigens and, therefore, cellular adhesion and signaling [18–21]. This process has been demonstrated previously in human oral cancers in which loss or aberration of ABO glycosyltransferase activity is thought to be an early event in carcinogenesis .
Finally, there is a strong association between soluble intercellular adhesion molecule-1 (sICAM-1) and the rs507666 single nucleotide polymorphism at the ABO locus that may explain the association between blood type and survival in patients with RCC . Levels of soluble cell adhesion molecules, e.g. sICAM-1, are known to be elevated in a number of human malignancies and may play a role in escape from immune surveillance by tumour cells . sICAM-1 has also been proposed as a biomarker in malignant melanoma and lung cancer [25,26]. Particularly relevant to the present findings of decreased survival in patients with non-O blood type, variance at one of the blood group A alleles was found to account for the association between levels of sICAM-1 and the ABO gene locus .
There are some important limitations to this single-centre retrospective study. Complete information allowing for the multivariable analysis was only available in 567 patients. Additionally, unidentified confounders may have biased our results. However, we controlled for a large number of known prognostic variables and we are unaware of any variables not included in the model that may be associated with both blood group and survival, a necessary attribute of any confounder. The distribution of ABO blood types in our patient cohort was random across all variables with exception of race, which had a distribution consistent with that reported in the literature . It is also important to note that a statistically significant association between blood type and OS was not observed in the univariable analysis, in contrast to the findings of the multivariable analysis. When constructing multivariable models, variables suspected to be associated with the outcome are included even when the univariable association does not reach statistical significance. This is because confounding effects may mask an association in the univariable analysis that is then observed in the multivariable analysis – in this case, an association between blood group and survival. Furthermore, although DSS approached significance, it was not significantly associated with blood group. It is therefore possible that blood type simply predisposes to an unrelated condition and does not directly affect or predict RCC recurrence or progression. These findings are contrasted, however, in the exploratory multivariable analysis in which individual blood types were included in the model. Blood type AB was found to be significantly associated with DSS, although this model may be overfit. Finally, these findings need to be confirmed in a prospective manner and with further understanding of the underlying biological mechanism.
In conclusion, we found a significant and independent association between ABO blood group and OS in patients undergoing radical or partial nephrectomy for locoregional RCC. Specifically, we identified non-O blood type as a predictor of mortality in these patients. ABO blood group has not been previously recognized as a predictor of outcomes after surgery for RCC, and it will be important to assess whether this adds to current preoperative prognostic nomograms. This finding may also have implications on a molecular level regarding the biology of RCC recurrence and progression. Further work to confirm these findings and understand the underlying biology may be important not only for RCC but possibly for other malignancies as well.
The project described was supported in part by Award Numbers K08 CA113452 (PEC) from the National Institutes of Health
CONFLICT OF INTEREST
Michael S. Cookson is a Paid Consultant for Endo, Myriad, and Spectrum. S. Duke Herrell is a Paid Consultant for Aesculap Inc, Covidien Surgical Devices, and Galil Medical, an investigator for Wilex, and a stockholder in Veran Medical Tech. Sam S. Chang is a Paid Consultant for Endo, Allergan, Amgen, and GE healthcare and has research funding through Endo. Peter E. Clark is a Paid Consultant for Galil Medical. Daniel A. Barocas is a Paid Consultant for Allergan and Dendreon.