Predicting donor specific crossmatch outcome using an in-house generated virtual crossmatch algorithm

详细信息    查看全文

• 作者：Prakash Rao ; Cecil Rhodes ; Donna King ; Dayanand Deo ; Patricia Harris ; Julien Napoleon ; Ijeoma Okere
• 刊名：Human Immunology
• 出版年：2015
• 出版时间：October 2015
• 年：2015
• 卷：76
• 期：supp_S
• 页码：144
• 全文大小：236 K

文摘

The new Kidney Allocation Scheme (KAS) has increased requests for virtual crossmatches (VXM) to predict the possibility of a negative donor specific cross-match. This study evaluates the efficacy of an in-house generated algorithm to predict and quantify the allogeneic cross-match.

Method

A retrospective review was conducted on 50 Flow Cytometry Crossmatch (FLXM) results that were collected from the database reflecting several different median channel shift (MCS) values. The corresponding HLA Class I and Class II antibody screening was performed using FlowPRA screen beads (One Lambda). Luminex Class I and Class II single antigen bead (SAB) testing and analyses including for mean fluorescent intensity (MFI) was performed. A calibration curve for predicting the VXM was developed by comparing the MFI values for the donor specific antibodies (DSAs) obtained from the Luminex SAB results to the corresponding MCS values from the FLXM results. A regression was performed on the MCS versus MFI values and the slope was calculated.

Results

We reviewed 246 FLXM results that were positive for T-cell and/or B-cell, performed between August 2008 and February 2013. We substituted the highest MFI values from the database for either the HLA class I or class II specificity in the VXM algorithm to obtain the predicted MCS values. We observed that our in-house generated VXM algorithm had an overall positive predictive value of 91% for the T-cell and B-cell crossmatches. MFI values above 1200 predicted a positive cross-match for HLA class I specificities while MFI values above 3000 predicted a positive crossmatch for HLA class II specificities. There were about 9% crossmatches that were predicted as negative using our VXM algorithm that were positive by FLXM evaluation. This could be due to Prozone like phenomenon, antibodies dependent on the conformational characteristics of the antigen, antibody avidity and affinity, or allele specificities not on the Luminex SAB panel.

Conclusion

The VXM algorithm shows a high probability of accurately predicting FLXM results, and may mimic the sensitivity of the final crossmatch reducing the time and cost for deceased donor work-up. This could allow for quicker organ allocation for sensitized patients.