Blood group distribution and life-expectancy: a single-centre experience

Introduction

The antigens of the ABO blood group system (A, B and H determinants, respectively) are complex carbohydrate molecules on the extracellular surface of red blood cells¹. However, besides being expressed on red blood cells, ABO antigens are also strongly expressed on the surface of a variety of human cells and tissues, including the epithelium, sensory neurones, platelets, and the vascular endothelium². The clinical significance of the ABO blood group system does, therefore, extend beyond transfusion medicine as several reports have suggested an important involvement in the development of cardiovascular, oncological and other diseases^3,4. In particular, a number of clinical studies and systematic reviews have documented a positive association between non-O blood type and the risk of developing both venous and arterial thrombotic events^5–8. Although the underlying molecular mechanisms are not completely understood, ABO-related thrombosis is thought to be mediated by ABO carbohydrate modifications of von Willebrand factor (VWF) resulting in impaired proteolysis and higher levels of VWF, and consequently higher factor VIII plasma levels in individuals with a non-O blood type than in blood group O individuals⁹.

Considering this association along with the finding that cardiovascular mortality is the first cause of death in men, it is not surprising that investigators have assessed whether ABO blood group correlates with life-expectancy^10–12. However, as the investigations gave conflicting results, we decided to analyse such an association at our city hospital.

Material and methods

Population data

A retrospective review of electronic clinical records was conducted for all subjects (outpatients and blood donors) who underwent ABO blood typing during the period from January 2010 to December 2013 at the Department of Haematology and Transfusion Medicine of Mantua. In all individuals evaluated, ABO blood type, age and gender were recorded. Age at the time of the observation was discretized into 11 age groups (0 to 10): age between 0–9, 10–19, 20–29, 30–39, 40–49, 50–59, 60–69, 70–79, 80–89, 90–99, and >99 years.

Statistical analysis

The mean age in females and males was compared by means of Student’s t-test. The correlation between blood groups and gender was evaluated using Pearson’s chi squared test on each of the O, A, AB and B groups. The changes of the frequencies of blood groups in various age groups were assessed with a test for trend, splitting the population by gender. The outcome was the odds of each blood group, where odds=P/(1−P), and P is the proportion of each blood group. The test computed a chi-squared test of homogeneity of odds (equal odds) and a test for linear trend of the log odds against the numerical code used for the categories of age. This investigation was applied to subjects between age 20 and age 99, comprising 26,788 observations. The same matter was investigated with a logistic approach, using each blood group frequency as a dependent variable, focusing on age as the main predictor, and adjusting for gender. The interaction age*gender was also added to the covariate list. The above age limits were maintained. All estimates were computed with Stata 13.1 software (StataCorp, College Station, TX, USA).

Results

During the four-year study period, a cohort of 28,129 subjects presenting at our Institution as outpatients or as blood donors was collected. Of these subjects, 20,897 were female and 7,231 were males. The summary data of the population are reported in Table I. The subjects’ ages spanned from 0 to 103 years. The mean age in females was 40.8 years whereas that in males was 57.2 years (p<0.0001). The age distribution showed a bimodal pattern both in females and in males (Figure 1). Comparison of the pattern of age distributions showed a female excess in the childbearing age.

Table I.

Characteristics of the population analysed, stratified by decades of age.

		Females		Males
Age	Group	Number	%	Number	%
<10	O	81	34.91%	125	42.09%
	A	102	43.97%	123	41.41%
	AB	17	7.33%	15	5.05%
	B	32	13.79%	34	11.45%
	Total	232		297
10–19	O	206	34.51%	83	40.69%
	A	246	41.21%	90	44.12%
	AB	51	8.54%	10	4.90%
	B	94	15.75%	21	10.29%
	Total	597		204
20–29	O	2,050	41.73%	127	37.13%
	A	1,700	34.60%	151	44.15%
	AB	287	5.84%	14	4.09%
	B	876	17.83%	50	14.62%
	Total	4,913		342
30–39	O	3,280	41.62%	312	42.80%
	A	3,223	40.90%	306	41.98%
	AB	325	4.12%	26	3.57%
	B	1,052	13.35%	85	11.66%
	Total	7,880		729
40–49	O	1,006	42.13%	304	41.03%
	A	1,008	42.21%	308	41.57%
	AB	101	4.23%	32	4.32%
	B	273	11.43%	97	13.09%
	Total	2,388		741
50–59	O	434	41.65%	343	40.16%
	A	458	43.95%	371	43.44%
	AB	43	4.13%	42	4.92%
	B	107	10.27%	98	11.48%
	Total	1,042		854
60–69	0	515	42.28%	622	41.00%
	A	545	44.75%	679	44.76%
	AB	38	3.12%	67	4.42%
	B	120	9.85%	149	9.82%
	Total	1,218		1,517
70–79	O	588	39.46%	715	42.66%
	A	693	46.51%	725	43.26%
	AB	56	3.76%	68	4.06%
	B	153	10.27%	168	10.02%
	Total	1,490		1,676
80–89	O	422	43.60%	331	42.01%
	A	416	42.98%	349	44.29%
	AB	38	3.93%	41	5.20%
	B	92	9.50%	67	8.50%
	Total	968		788
90–99	O	66	40.99%	37	45.68%
	A	70	43.48%	36	44.44%
	AB	8	4.97%	3	3.70%
	B	17	10.56%	5	6.17%
	Total	161		81
>99	O	6	75.00%	1	50.00%
	A	2	25.00%	0	0.00%
	AB	0	0.00%	1	50.00%
	B	0	0.00%	0	0.00%
	Total	8		2
Total	O	8,654	41.41%	3,000	41.49%
	A	8,463	40.50%	3,138	43.40%
	AB	964	4.61%	319	4.41%
	B	2,816	13.48%	774	10.70%
	Total	20,897		7,231

Figure 1.

Age distribution in females (F) and males (M).

A significant association between male gender and blood group A was found (frequency: 43.40% in males, 40.50% in females; p<0.001). A similar, but inverted, association was found for group B (frequency: 10.70% in males, 13.48% in females; p<0.001). No significant differences related to gender were detected for groups O and AB. The test for trend with age produced the inferential estimates reported in Table II. In the female population, the group A frequency, measured as odds, exhibited a significant trend to increase with age, whereas the frequencies of groups AB and B manifested a significant trend to decline (Figure 2a). In the male population, only group B exhibited a significant trend to change (decrease) with age (Figure 2b). Interestingly, seven out of the ten subjects aged >99 years had blood group O, although this result did not reach statistical significance (Fisher’s exact test). Using a logistic approach, age had an increasing effect on group A, decreasing on groups AB and B, with significant differences between the two genders for groups A and AB, as shown in Figure 3.

Table II.

Blood group prevalence trend with age according to gender.

		Group O	Group A	Group AB	Group B
Females	Homogeneity	NS	p<0.0001	p<0.0001	p<0.0001
	Trend	NS	p<0.0001	p=0.0002	p<0.0001
Males	Homogeneity	NS	NS	NS	p=0.0084
	Trend	NS	NS	NS	p=0.0001

In females, group A showed a significant increasing trend, while groups AB and B showed a significant decreasing trend with age.

In males, group B showed a significant decreasing trend with age.

NS: not significant.

Figure 2.

Changes in blood group frequencies according to age and gender.

Figure 3.

Predictions of blood group probabilities by age and gender: logistic regression.

“z” is the ratio between the regression coefficient and the corresponding standard error; from “z” stems the probability “p” of coefficient=zero. The age had an increasing effect on group A, decreasing on groups AB and B. The differences between the two genders were significant for groups A and AB.

F: females; M: males.

Discussion

It has been known for many years that the ABO blood type has a profound influence on haemostasis, being a major determinant of VWF and, consequently, of factor VIII plasma levels^13,14. In particular, VWF levels are approximately 25% higher in individuals who have any blood group other than O¹⁴. As VWF and factor VIII are well-established risk factors for thrombosis^15,16, it is understandable that a number of experimental and clinical studies have assessed whether ABO blood type could influence the risk of developing arterial or venous thrombotic events and, ultimately, human longevity¹⁷. To verify this latter issue, we performed a large retrospective survey on more than 28,000 individuals stratifying their blood group according to age and gender. Interestingly, we observed that while the number of subjects with group A increased with age, the numbers with group AB and B decreased, indicating a negative association between these blood types and longevity. Interestingly, our findings are partially in accordance with those published by Brecher and colleagues¹¹, whose study design was similar to that utilised in our research. Indeed, the authors retrospectively reviewed blood group distribution in a cohort of patients stratified by decade of death and found that patients with group B blood had an overall decreased survival (p<0.01) compared with patients with the other blood groups. However, the results of our investigation and that by Brecher and colleagues are not in accordance with those from other studies. For instance, Shimizu and colleagues¹⁰ analysed the ABO blood type frequency in 269 centenarians and 7,153 controls and found that blood type B was statistically more frequent among centenarians than among controls (29.4% vs 21.9%, p=0.04). No association between ABO blood group and longevity was observed by Vasto and colleagues¹² in a study on 38 centenarians and 59 healthy controls. Differences in the design of the various studies (the studies by Shimizu and colleagues¹⁰ and Vasto and colleagues¹² were focused exclusively on extremely old subjects) may account for the discrepancies of the findings. Interestingly, a high proportion of centenarians in our cohort (70%) were of O blood type, which has been found in several studies and meta-analyses to protect from cardiovascular diseases and cancers (e.g., pancreatic and gastric cancers)¹⁷.

Conclusion

In conclusion, our retrospective survey showed that the percentage of people with group B blood declined with age. Group AB also had a negative correlation with age, although this was less pronounced: indeed, its effects were conditioned by gender, being significant only in females. The proportion of subjects with group A blood increased with age, but again this effect was significant only in females. Thus a conditioning effect of gender was evident for both A and AB groups. We have no explanations for these observations, although an association between B blood type and some aging-associated degenerative disorders, such as Parkinson’s disease, has been found¹⁸.

We are aware of the possibility of multiple potential confounding effects, which might be active in this study, but not explicitly included in the evaluation. This is an observational study of the population assisted by a Department of Haematology and Transfusion Medicine. Young adults are more often blood donors or pregnant women undergoing Rhesus incompatibility assessment; old subjects are more often patients. Thus, disease status might be associated with blood groups in a different manner in different age groups. Moreover, information on the ethnic background is missing; this variable could be related both to age and blood groups. Nevertheless, we believe that the present investigation could be viewed as an initial attempt to shed light on a topic deserving attention. Further studies on large populations of patients, with particular effort to eliminate potential confounding effects, are necessary to verify our preliminary findings.