In a Mar 2010 blog post, I discussed the long transplant wait times faced by kidney patients in the O blood group. In that post, I made some simplifying assumptions regarding the distribution of blood groups. I also ignored incompatibilities due to human leukocyte antigens (HLA). The complete list of seven assumptions are shown in the footnote under Table 1 of that blog post.

I want to eliminate those simplifications and do a deeper analysis. In today’s post, I will start by checking to see if there is any correlation among kidney patients on the UNOS transplant waiting list between ABO blood group and the likelihood that the patient has antibodies to foreign HLAs. In general, the recipient must receive a donor organ for which they do not have antibodies against.

It is believed that people are not born with antibodies to foreign HLAs. They become sensitized to foreign HLAs due to blood transfusions, previous transplants, or pregnancies (May 2008 blog post). These antibodies will cause them to reject organs that are not compatible. One measure of the levels of HLA antibodies a person carries is called calculated panel reactive antibodies (cPRA). Calculated PRA indicates the percent of donors in the U.S. (based on historical U.S. deceased donor data) whose HLAs would react with the patient’s antibodies. If the patient has no HLA antibodies, then the cPRA is 0%. As the number of antibodies in the patient rises, the percentage of donor organs that don’t match rises. Not all HLAs are equally common and that affects the calculated PRA result as well. Any patient with a cPRA of 80% or higher is considered hard-to-match and is accorded extra points to put them closer to the top of the waiting list. The cPRA of patients on the waiting list can rise over time as they get additional transfusions. It can also fall, as HLA antibodies disappear in a process that isn’t well understood.

In order to investigate the relationship between blood type, HLA, and other factors, we need to know if these variables are correlated. Specifically, are people with a particular blood group more likely to have high cPRA percentages? If these two variables are correlated (not independent), then any further data analysis becomes much more complex. In order to test this, I requested some data from UNOS. I want to thank Bruce Shepperson of UNOS for generating this data for me. Any errors in analysis are mine alone.

The cPRA is not reported for all patients. To make data comparable, in the tables below I added a row that excludes the Not Reported patients and calculate an adjusted total. (Note: This exclusion assumes that there is no bias in the Not Reported group. But this may be incorrect. Perhaps transplant centers will be more likely to measure the cPRA of patients whose medical history includes previous blood transfusions, transplants, or pregnancies. If so, then the Not Reported column is more likely to contain patients with low cPRA. In fact, starting on Oct 1, 2009, the UNOS assumes any patients that do not have reported HLA sensitivities have a cPRA of 0%.)

In the tables below, cells with higher than expected counts (at the 0.05 significance level) are shaded green while those with lower than expected counts are shaded yellow. In the rows and columns containing totals, the cells are compared to the corresponding cell in Table 1.

Patients added to the waiting list during a year

The first test is to see if the distribution of blood groups and cPRA has changed over time for patients added to the active and inactive kidney-only waiting lists. An explanation of these waiting lists in a Apr 2010 blog post. Table 1, shows the distribution for patients added in 1995. Patients in the O blood group are more likely to have 1-19% cPRA than expected, but I suspect this is just random noise. (At the 0.05 significance level, you would expect 1 out of 20 significant results to be due to noise.)

Table 1. Blood group by cPRA* for patients added to waiting list in 1995 (n=17,872)

 

0%

1-19%

20-79%

80+%

Not Rpt’d

Row Total

O

24.5%

7.7%

3.7%

1.5%

10.3%

47.7%

A

18.5%

4.8%

2.2%

1.0%

7.5%

34.0%

B

7.7%

2.0%

0.9%

0.5%

3.5%

14.5%

AB

2.1%

0.5%

0.3%

0.1%

0.8%

3.8%

Column Total

52.8%

15.0%

7.0%

3.1%

22.0%

100.0%

Adjusted Total

67.8%

19.2%

9.0%

4.0%

100.0%

  *Measured at time of listing

Table 2 shows the distribution of new patients added to the waiting list 14 years later, in 2009. None of the cells in Table 2 show significant differences. Notice though that the proportion of patients with 0% PRA is higher than in 1995. There may be several explanations for this. First, eliminating the use of human-derived EPO for dialysis patients eliminates that as a source of foreign HLA. Perhaps changes in the way cPRA is calculated over this time (as DNA testing becomes more accurate) may produce more 0% results. Also note the number of new patients nearly doubled from 17,872 to 35,122 over this period. So the proportion of first time transplant patients to repeat patients is probably rising. New patients are probably more likely to have a cPRA of 0%. The incidence of end-stage renal disease is growing very fast. It really is a health care crisis.

Table 2. Blood group by cPRA* for patients added to waiting list in 2009 (n=35,122)

 

0%

1-19%

20-79%

80+%

Not Rpt’d**

Row Total

O

26.1%

3.8%

3.2%

2.0%

13.4%

48.5%

A

17.8%

2.6%

2.2%

1.5%

8.9%

33.0%

B

7.8%

1.2%

1.0%

0.6%

4.0%

14.6%

AB

2.1%

0.3%

0.3%

0.2%

1.0%

3.9%

Column Total

53.9%

7.9%

6.7%

4.2%

27.3%

100.0%

Adjusted Total

74.2%

10.8%

9.2%

5.8%

100.0%

*Measured at time of listing
**
As of 10/1/2009, if no unacceptable antigens are reported, cPRA value defaults to 0%

Overall, we can conclude that there is no correlation between blood group and cPRA for patients entering the UNOS waiting list.

Patients taken off the waiting list during a year

The next test is to see if there is a correlation between blood groups and cPRA among patients who received a transplant. Tables 3 and 4 show the distributions for patients receiving a kidney during 2009 from a live donor and deceased donor respectively.

Table 3 shows no discernible relationship between blood group and cPRA. There is a relationship in Table 4. Among patients receiving a deceased donor transplant, those with O blood and high cPRA are more likely than other blood groups with high cPRA. This is a result of the UNOS allocation rules that favor this combination and so is expected. These patients are the hardest to match, so they are given extra points which helps gets them off the list faster than expected by chance.

Comparing the totals in Table 3 and 4 against the totals in Table 1 shows that patients with type O blood, who can only accept a type O kidney, are less likely to get a transplant than patients with type A or AB. This is expected. (Patients with type B blood are also less likely to get a transplant, but that is because of different issues which I will discuss in a future blog post.)

But there is an unexpected result. Patients with high cPRA are more frequent among those receiving a kidney transplant than those entering the waiting list. This seems counterintuitive since they are harder to match. There are four possible explanations. First, maybe patients with high cPRA are exiting the list faster than patients with low cPRA. We know this is not true, and Table 5 (which we will look at shortly) shows it.

A second explanation requires us to notice the low proportion of Not Reported for patients receiving a transplant from a deceased donor. Perhaps most of the Not Reported in Tables 1 and 2 have high cPRA, which is the opposite of what I posited earlier. Third, cPRA is not static like blood type. Maybe it rises between the time the patient is added to the waiting list to the time of the transplant. Assuming no bias in the Not Reported group, this means about a fourth of all patients have their cPRA score rise significantly during their years on the waiting list. Finally, perhaps the cPRA data collected at the time the patient entered the list was incorrectly low and was updated to a higher value by the time of transplant.

The discrepancy between cPRA for patients entering the waiting list and those leaving it is an issue worth exploring. But it does not affect our conclusion that there is no unexpected correlation between blood type and cPRA among those receiving a transplant.

Table 3. Blood group by cPRA* for patients receiving a live donor transplant during 2009 (n=6,387)

 

0%

1-19%

20-79%

80+%

Not Rpt’d**

Row Total

O

23.7%

5.9%

4.7%

1.6%

9.2%

45.0%

A

20.6%

4.4%

3.4%

1.3%

8.2%

37.9%

B

6.9%

1.4%

1.6%

0.5%

2.8%

13.2%

AB

2.1%

0.5%

0.3%

0.1%

0.9%

3.8%

Column Total

53.3%

12.2%

10.0%

3.4%

21.1%

100.0%

Adjusted Total

67.6%

15.4%

12.7%

4.3%

100.0%

*Measured at time of transplant

**As of 10/1/2009, if no unacceptable antigens are reported, cPRA value defaults to 0%

Table 4. Blood group by cPRA* for patients receiving a deceased donor transplant during 2009 (n=10,442)

 

0%

1-19%

20-79%

80+%

Not Rpt’d**

Row Total

O

24.4%

6.0%

6.4%

7.1%

1.4%

45.3%

A

20.0%

5.1%

5.4%

5.1%

1.0%

36.5%

B

7.1%

1.9%

2.0%

1.4%

0.4%

12.9%

AB

3.1%

0.8%

0.8%

0.6%

0.1%

5.3%

Column Total

54.6%

13.8%

14.6%

14.2%

2.9%

100.0%

Adjusted Total

56.2%

14.2%

15.0%

14.7%

100.0%

*Measured at time of transplant
**As of 10/1/2009, if no unacceptable antigens are reported, cPRA value defaults to 0%

Patients remaining on waiting list at end of a year

The final test is to see if there is a correlation between blood group and cPRA among patients who remain on the waiting list at the end of a year. The data in Table 5 below shows the distribution for the 57,203 patients on the active kidney-only waiting list. Looking at the row totals, as expected, patients with type O blood, who are harder to match, are more likely to be on the waiting list at the end of the year than patients with types A and AB. Similarly, looking at the adjusted column totals, patients with high cPRA are more prevalent than those with low cPRA.

Notice that there is a correlation between blood type and cPRA in this table that is the opposite of Table 4. Among patients with high cPRA, type AB blood group is more prevalent than expected and type O blood group is less prevalent. This is a direct result of the allocation scheme that UNOS has developed to favor patients with type O blood and high cPRA. Among patients with 0% cPRA, those with type AB blood group can accept a kidney from almost anybody, so they are less likely to remain on the list at the end of the year, while those with hard to match type O blood group are more likely to remain on the list.

Table 5. Blood group by cPRA* for patients on waiting list at end of year 2009 (n=57,203)

 

0%

1-19%

20-79%

80+%

Not Rpt’d**

Row Total

O

34.4%

3.4%

7.1%

8.6%

0.0%

53.4%

A

17.1%

1.5%

3.5%

5.1%

0.0%

27.2%

B

10.4%

1.1%

2.4%

2.8%

0.0%

16.7%

AB

1.6%

0.2%

0.4%

0.6%

0.0%

2.7%

Column Total

63.4%

6.1%

13.4%

17.0%

0.0%

100.0%

Adjusted Total

63.4%

6.1%

13.4%

17.0%

100.0%

*Results of latest available test
*
*As of 10/1/2009, if no unacceptable antigens are reported, cPRA value defaults to 0%

Thus, we can conclude that there is no unexpected correlation between patients’ blood type group and their cPRA. However, allocation rules and match probabilities will alter the composition of the waiting list.

[Update: Due to an editing error, some of the cells were shaded incorrectly. The corrected data is now shown. The changes do not affect the conclusion that there are no unexpected correlations between blood type and cPRA.]

Advertisements