Blood Type Distribution in Autoimmune Diseases: An Anonymous, Large-Scale, Self-Report Pilot Study

Background: Recent research has veried that blood group or Rh factor can inuence susceptibility to various cardiovascular, neoplastic and infectious diseases including COVID-19. While a number of studies have looked at correlations between blood group and various rheumatological diseases, ndings have been inconsistent, often because many of these studies suffered from small sample size issues. In order to better understand the potential relationships between blood group/Rh factor and rheumatological diseases, we performed a large-scale self-report pilot study of blood type distributions in ve autoimmune diseases. Methods: Five autoimmune diseases were included in the study: systemic sclerosis, systemic lupus erythematosus, rheumatoid arthritis, psoriasis, and ankylosing spondylitis. We also included a control group in which participants did not have any autoimmune diseases. The participants were recruited through social media and organizations such as the Lupus Foundation and the National Psoriasis Foundation. Respondents who met the inclusion criteria were asked only two questions by anonymous survey: blood type and country of birth. Results: Each autoimmune disorder group included between 570 and 951 US participants. While there was little difference in blood type distribution patterns among the ve diseases, unexpectedly, all ve disease groups showed a consistent pattern where Rh negative was almost twice as high as US population norms. A post-hoc non-autoimmune control group was added in order to determine if this anomalous nding was an artifact of the study design. The control group displayed a similar unexpected increase in the Rh-negative blood type prevalence, suggesting that the very high Rh-negative frequency among the tested disease groups was likely to be an artifact of the study design. Conclusions: Overall, our preliminary study results show no meaningful differences between the disease groups and the post-hoc control group, suggesting that neither ABO type nor Rh factor affects susceptibility to the development of any of the ve studied autoimmune diseases. Nevertheless, the unexpected observed difference in Rh factor distribution between the studied groups/control group and the corresponding US population norms has important implications for any research study using self-selected subjects. Our results suggest that such studies may be subject to unanticipated biases,


Background
Over the past 100 years, hundreds of studies have been done looking at statistical relationships between blood groups and medical conditions or traits such as personality and even criminality. While many of these early studies were controversial and suffered from numerous methodological aws, including small sample size and incorrect analysis, more recent research has demonstrated that there are clear associations between blood groups or Rh factor and susceptibility to various diseases, including infectious and, cardiovascular diseases, certain cancers, and most recently, COVID-19. (1)(2)(3)(4)(5)(6) A number of studies have looked at distribution patterns of ABO group and Rh factor in multiple autoimmune diseases, including rheumatoid arthritis (RA), ankylosing spondylitis (AS), systemic lupus erythematosus (SLE), systemic sclerosis (SSc), and Sjögren's syndrome (SjS). (7)(8)(9)(10)(11) In general, most of these studies found little or no difference in ABO group distribution patterns in the various studied diseases, but several studies did report differences in Rh factor between study groups and control groups in several autoimmune diseases, but even there, ndings were inconsistent in the various studies. A likely reason for the disparity in results is that in several of these studies, group sizes were quite small.
There is a perception in the popular press and in patient support groups that patients with Rh negative blood are more susceptible to autoimmune diseases. In 2016, the Scleroderma Education Project (a 501c3 non-pro t organization focused on systemic scleroderma education and research) conducted a large-scale, self-report online survey of SSc patients to see if there was a basis for conducting a formal study to investigate blood type distribution patterns in patients with SSc. This initial survey of 743 respondents found that Rh-negative SSc patients were signi cantly more common than was expected. Unfortunately, this survey failed to assess country of birth, making it impossible to do a valid statistical analysis since different countries have signi cantly different blood-type distribution norms.
In April 2017, we decided to do a formal large-scale self-report pilot study of blood type distributions in ve separate autoimmune diseases in order to see if a formal follow-up study using patient records could potentially be justi ed.
There is inherent concern with self-report data because of potential issues like self-selection biases and self-report accuracy. However, previous research has shown that there is good to very good agreement between self-report and medical records for both diagnosis and symptoms (12)(13)(14), suggesting that a self-report study is an appropriate rst step before initiating a more expensive formal study using medical records.
The ve diseases chosen for the study are listed below: SSc SLE RA Psoriasis AS SSc, SLE, and RA are strongly female dominant connective tissue diseases that are usually positive for anti-nuclear antibodies (ANA). Psoriasis is also female dominant but is ANA negative. AS is one of the rare autoimmune diseases that is male dominant.
Initial goals for the study were to determine 1) if there were any signi cant differences in blood type distribution patterns among the ve surveyed diseases, and 2) if the blood type distribution patterns for any of these diseases differed from the US population norms. Our primary focus was to determine if clear patterns were evident that would justify a more formal follow-up study rather than on detailed quantitative analyses of the data. This study was determined to qualify as an exempt study by the University of Wisconsin (Madison) Institutional Review Board (Project No: 2017-0545).

Autoimmune Disease Blood Type Survey Design
The survey was restricted to just two questions: What is your blood type?
What is your country of birth?
Response choices consisted of a list of 100 countries. The rst three were United States, United Kingdom, and Canada. Following these three countries, additional countries listed were alphabetical. There was also an "Other (please specify)" option that allowed someone to enter the name of any country that was not on the list.
The surveys were done using a common online survey tool called SurveyMonkey. The surveys were con gured so that only one survey was allowed per IP address. While technically, anyone could take the survey more than once, this would require them to take the survey again from a different physical location or to use a software tool like a Virtual Private Network to alter the normal IP address, which we believe to be unlikely to occur given the nature of the surveys. Table 1 shows the speci c inclusion/exclusion criteria listed for each of the ve individual surveys. Patients need to know their blood type to participate in the surveys. Our target goal was a minimum of 500 respondents per disease with the US as country of birth. Recruitment was done primarily through social media, for example through postings in individual disease-focused patient support groups on Facebook or other organizations. The surveys were also announced in online news websites such as Scleroderma News and Lupus News. Several organizations, including the Lupus Foundation and the National Psoriasis Foundation, also publicized the study. The rst survey was launched in February 2017 and the nal survey was closed in November 2017.
Non-Autoimmune Control Group * In order to properly analyze the results of our survey of patients with autoimmune diseases, we included a control group of people that did not have any type of autoimmune disease. Control group participants could not be either: 1) formally diagnosed with, or 2) in the process of being evaluated for any autoimmune disease, not limited to the ve diseases we were studying, but also including other autoimmune diseases such as multiple sclerosis, Hashimoto's disease, or SjS. Control group survey participants were recruited through groups and methods that had no disease association, including non-medical Facebook groups and also through Amazon MTurk.
*Note: the control group was added post-hoc after a preliminary analysis of the data showed unexpected distribution patterns.

US Blood Type Distribution
Blood type distribution norms for the US are set out in Table 2.

Results
We were able to easily achieve our overall goal of a minimum of 500 US survey participants per disease. Table 3 below shows the total number of survey responses per disease and the number of surveys where the selected country of birth was the US.  Table 4 shows the detailed breakdown of blood types for each of the ve diseases as well as the US population norms from the Stanford School of Medicine Blood Center (Table 2). Table 5 shows the result breakdown by ABO Type. Table 6 shows the result breakdown by Rh Factor.

Statistical Analysis
There were two questions these questionnaires were designed to address: 1) Does blood type distribution differ among these different autoimmune diseases?

Discussion
While the main goals for this preliminary study were to look at blood type distribution patterns in ve different autoimmune diseases compared to US population norms and each other, our most striking nding was that the bloodtype distribution pattern in our control group of people without any association with any autoimmune disease was substantially different from the expected population norms. While there are some statistically signi cant differences in distribution patterns among the ve diseases, overall, the ve diseases showed similar distribution patterns, and all ve diseases were very similar to the control group.
If only the ABO distribution is analyzed (with no consideration of Rh factor), the ve diseases and the control group were relatively close to the expected US population norms for ABO type distribution, with the exception of the rarest ABO type -AB. While this may be an anomaly due to relatively small numbers of this least frequent blood type, there may be another explanation for this consistent nding across all disease groups plus the control group. This is discussed below.
However, the most striking nding was the unexpected difference between the Rh-positive and Rh-negative individuals in all studied groups. While there were few signi cant differences among the ve disease groups or between the control group and the disease groups, there was a very large difference in Rh factor distribution (positive versus negative) between the Stanford US population data and both the disease group data and the control group data, with the percentage of respondents who were Rh-negative being 70% to 95% higher than expected depending on the speci c disease.

Understanding the Results
There are two factors that need to be considered that might account for some or all of the anomalous survey results:

Random Guessing
Some of the people taking the survey are probably wrong about their blood type. If we assume that the incorrect blood type "guesses" are evenly distributed among the eight possibilities, the effect of random guessing is to increase the Answer: Yes: Χ 2 (28) = 48.1, p < 0.01 2) Is the distribution of blood types in each of these ve autoimmune diseases the same as in the general population?
The answer to this question is more complicated. When we saw that the results were vastly different from the general population, we added a control group to determine if the unexpected results were an artifact of the experimental design or an actual re ection of blood type distributions in our studied disease populations.
As is readily visible in Fig. 1, all ve of the disease groups and the control group are highly signi cantly different than published population norms (p < 0.0001). In looking at differences between individual diseases and the control group (Table 7), only SSc, RA, and AS do show statistically signi cant differences between these diseases and the control group. p = 0.053 expected frequencies of any rare blood type (e.g., O-) and decrease the expected frequency of any common blood type (e.g., A+). While this type of error may potentially account for a small portion of the anomalous results, it does not explain why the Rh-, and in particular O-, observed results are much higher than expected.
A "Red Cross Effect"?
A second, and much more likely explanation for the unexpected results is a well-documented phenomenon called the "availability heuristic" or "availability bias" (15). An availability heuristic is a type of mental shortcut that involves basing judgements on information and examples that immediately spring to mind.
Most people in the US learn their blood type either as part of a medical procedure or when donating blood or blood products. When people desire to donate blood or blood products at the Red Cross or other organization, their blood is screened at the initial visit for potential transmissible diseases, and it is also typed. If it turns out that they are Onegative in particular, they will be informed that they are "universal donors" and that their blood is highly sought after by organizations like the Red Cross.
While this needs to be con rmed by research comparing known blood types through medical records versus self-recall of blood type, it is possible that being told that one's blood type is in high demand may substantially increase the likelihood of remembering correctly the directly-determined blood type. Also, blood drives often emphasize getting blood donations from people who are Rh-negative since Rh-negative blood can be transfused into many more people than Rhpositive blood.
Note: In looking at the blood type distribution patterns for each disease group and in comparison to the control, there are a few statistically signi cant differences that may be worth examining in future research. Figures 2 and 3 suggest that these differences stem from ABO blood group rather than Rh factor considerations.

Conclusion
Our self-report pilot study suggests that there are no signi cant differences in blood type distribution among the ve surveyed autoimmune disease groups or between the disease groups and a post-hoc control group. This suggests that at least for these ve surveyed autoimmune diseases, there is no indication that either ABO type or Rh factor increases or decreases the likelihood of patients being diagnosed with any of these diseases. Based on this, there does not appear to be any justi cation for doing a follow-up study using actual patient records for any of these diseases. Certainly, our ndings do not exclude the possibility that the ABO type or Rh factor may play a role in some other autoimmune disease and any future research of this nature should consider the methodology limitations observed in our study.
Nevertheless, the large difference in Rh factor distribution (positive/negative) between the disease and control groups versus the Stanford US data norms is de nitely worthy of further investigation. Although there is no reason to doubt the accuracy of the Stanford data or similar charts from the American Association of Blood Banking, obviously there is some cause leading to a self-selection bias effect. While this could be from the availability heuristic such as our hypothesized "Red Cross Effect" or from random error, as discussed earlier, it could be due to some other, completely unknown, factor(s). Further research is needed to determine the reason for the anomalous results observed in our study.
As noted earlier, research studies suggest that self-reporting of diagnosis is usually in good agreement with the patient's medical record. However, it is likely that patients may be less accurate in determining whether or not they meet the exclusion criteria for each disease. This raises the question of whether or not some patients should have been excluded from participating in the survey, although this issues may be somewhat mitigated by the very large number of subjects surveyed in each disease group. Obviously, there are clear limitations in a preliminary study of this nature, some of edited the nal manuscript.