title page - research portal, king's college, london · web viewfrom the uk adult kidney...
TRANSCRIPT
Accepted for publication in The Lancet on 30th of August 2016. Ref: THELANCET-D-16-01100R2
Title PagePost-listing survival for highly sensitised patients on the UK kidney transplant waiting list, a matched cohort analysis
Authors: Manook M1, Koeser L2, Ahmed Z1, Robb M3, Johnson R3, Shaw O1, Kessaris N1, Dorling A1,4, Mamode N1
Department & Institution: 1. Department of Transplantation, Guy’s and St Thomas’ NHS Foundation Trust, 2. King’s Health Economics, Health Services and Population Research Department,
Institute of Psychiatry, Psychology and Neuroscience at King’s College, London3. NHS Blood & Transplant, Bristol, UK4. MRC Centre for Transplantation, King’s College London, Guy’s Hospital
Corresponding author:Professor Nizam MamodeDepartment of Transplantation6th Floor, Borough WingGreat Maze Pond, London, EnglandSE1 9RTTelephone: +44 (0)20 7188 1543Fax: + 44 (0)20 7188 5646Email: [email protected]
Running Title:Post-listing survival for highly sensitised patients on the UK kidney transplant waiting list
Key words:HLA-incompatible, Living Donor Renal Transplant, Highly Sensitised, Antibody incompatible transplantation, Survival.
Abbreviations:
ABOi ABO antibody incompatibleCDC Complement Dependent CytotoxicitycDD Compatible Deceased DonorcLD Compatible Living DonorDBD Donation after brainstem death
1
DCD Donation after circulatory deathDD Deceased Donor recipientDSA Donor Specific AntibodyFXCM Flowcytometric cross matchHLAi Human Leukocyte Antigen- incompatibleLD Living donorKPD Kidney Paired DonationUKLKDSS UK Living kidney donor sharing scheme
2
Abstract
Background: Over 40% of patients awaiting a kidney transplant in the UK are sensitised with HLA antibodies. Median time to transplantation for such patients is double that of unsensitised patients. Removing antibody to perform an HLA-incompatible (HLAi) living donor transplantation is perceived to be high risk, however, patient survival data is limited. We compared survival of patients opting for an HLAi kidney transplant to that of similarly sensitised patients awaiting a compatible organ.
Methods: From the UK adult kidney transplant waiting list, we selected crossmatch positive living donor HLAi kidney transplant recipients who received their transplant between 1.1.2007 and 31.12.2013, and were followed up to 31.12.2014 (end of study), along with a matched cohort of similarly sensitised patients listed for a deceased-donor transplant during that period. Data was censored both at the time of transplant (‘Listed only’), and at the end of the study period (‘Listed or Transplant’). We used Kaplan-Meier curves to compare patient survival between HLAi and the matched cohort.
Findings: Of 25,518 patient listings, 213 underwent an HLAi transplant during the study period. 852 (4 to 1) matched controls were identified. Of those, 41% remained untransplanted at just under 5 years after matching. There was no statistically significant difference in patient survival between ‘HLAi’ and either ‘Listed only’ (log rank, p = 0·446), or ‘Listed or Transplant’ (log rank, p = 0·984).
Interpretation: Survival of sensitised patients undergoing HLAi in the UK is comparable to those on dialysis awaiting a compatible organ, many of whom are unlikely to be transplanted. Choosing a direct HLAi transplant has no detrimental impact on survival, but conversely offers no survival benefit, in contrast to similar patients studied in a North American multicenter cohort.”
Funding: UK National Health Service Blood & Transplant, and Guy’s & St Thomas’ National Institute for Health Research Biomedical Research Centre.
3
Introduction
For many patients with end-stage renal disease (ESRD) awaiting a kidney transplant,
sensitization to Human Leukocyte Antigens (HLA) by blood transfusion, pregnancy or
previous transplantation, resulting in circulating antibodies (HLA-antibody), adversely
affects the chances of obtaining a suitable donor organ. In the UK, the level of sensitization
is estimated by the calculated reaction frequency (cRF), which estimates the percentage of
deceased donors against whom the recipient has preformed anti-HLA antibodies. Over a
quarter (26%) of patients on the waiting list are ‘highly sensitised’, with a cRF ≥85%; for
these patients transplant waiting time is doubled to a median of 6 years, compared to
patients with a cRF <10% (1).
Living donor kidney transplantation (LDKT) is the preferred treatment for ESRD, but for
many sensitised patients, this is impossible because anti-donor HLA antibodies result in a
positive crossmatch against their donor. This test indicates a very high risk of immediate
rejection(2) and therefore precludes transplantation, unless the patients undergo
antibody-removal (or ‘desensitisation’). There are five transplant possibilities currently
available for such patients: direct HLA-incompatible (HLAi) living donor kidney
transplantation following desensitisation; crossmatch negative (i.e.’compatible’) living
donor transplantation via the UK Living Donor Kidney Sharing Scheme (UKLDKSS); a
combination of both UKLDKSS and HLAi-transplantation; finding an alternative compatible
living donor, or awaiting a compatible deceased donor kidney.
In principle, the UKLDKSS offers a good opportunity for such patients to obtain a
compatible living donor transplant through a kidney paired donation (KPD) scheme. In
reality, the number of highly sensitised patients transplanted remains small (3). In 2014-
2015, for example, 66% of patients registered in the paired scheme had a cRF of >85%.
However, between 2012-2015, only 33% (n = 67) of UKLDKSS transplants were
undertaken in patients with a cRF >85%. Patients with a cRF of >95% have consistently
been found to make up 50% of the pool in each matching run (unpublished data, NHSBT).
Direct HLAi transplantation requires desensitization prior to transplantation (usually with
a combination of plasmapheresis and IVIg, to achieve a negative crossmatch) and enhanced
immunosuppression for induction and maintenance. This type of intensive regimen is not
4
suitable for patients with significant co-morbidities (4-7). In the UK, direct HLAi is not
currently offered by all centres.
For patients where direct HLAi transplantation is not feasible, either because of co-
morbidities or because the level of donor specific antibody (DSA) is too high to reduce, and
paired donation has not been successful, a combination of UKLDKSS and HLAi can be
used(8, 9), where desensitization allows a feasible HLAi transplant within the sharing
scheme. Unfortunately, for many sensitised patients these options are either not available
or acceptable, and the only other transplant options are to find alternative compatible
living donors, or wait for a compatible deceased donor kidney (10, 11); the wait may be
long or indefinite.
A key consideration for transplantable sensitised patients, is the likelihood of survival after
HLAi transplant. Desensitisation treatment itself confers increased risk of morbidity,
and HLAi results in an increased mortality and morbidity when compared with compatible
transplantation (12). In the USA, HLAi has been demonstrated to be more cost-effective
than dialysis (13).
Although this kind of data is useful for those patients who have a compatible donor, the
practical question for most patients and transplant doctors is whether to go ahead with a
direct HLAi transplant or wait on dialysis for one of the described alternatives, which might
never materialise. Multicentre data from the USA suggests that there is a survival benefit in
having an HLA-incompatible transplant, compared to waiting on dialysis for a compatible
alternative (14). However given the difference in dialysis mortality between USA and
Europe(15, 16), this may not hold true in the UK. Therefore, we used a national multicentre
assessment of survival outcome to answer the question, ‘Is the survival of highly sensitised
patients in the UK, following a direct HLAi transplant requiring desensitization, different to
patients who wait for a compatible alternative from the deceased donor list or through the
KPD scheme?’ In addition, for those patients choosing to await a compatible transplant, we
wanted to characterize how many patients were transplanted, and compare graft survival
between comparable patients receiving HLAi compared to compatible alternatives.
5
Methods
Study population
Patients were selected from the UK Adult renal transplant waiting list database for this
matched case control study. Managed by NHS Blood & Transplant (NHSBT), it contains
mandatory listing data used for renal transplant allocation. Information relating to
registration in the UKLDKSS was also linked using the registration ID. Data was extracted
relating to all patients on the waiting list on 1.1.2007, and all subsequently listed patients
until 31.12.2013, with follow up data to 31.12.2014 (n = 25,518).
The treatment group of interest were HLAi recipients of living donor renal transplants
from the UK renal transplant waiting list with a positive cross match after 1.1.2007. A
positive cross match was defined either by flow cross matching, (FXCM), or complement
dependent cytotoxicity (CDC), and was reported by local laboratories to the central
database. Although some centres in the UK classify cross-match negative recipients with a
DSA identified by Luminex testing as HLAi transplants, this is not consistent so these
patients are not included in this analysis. The controls were selected from patients listed
after 1.1.2007 who did not receive an HLAi transplant.
Patients originally listed on the paediatric waiting list, ABO-incompatible recipients and
deceased donor HLAi recipients were excluded, as these factors might be expected to be
associated with a different degree of risk when compared to adult living donor HLAi.
The primary outcome measure was post-transplant survival for HLAi, and post-listing
survival for the control groups. Further, we provide a detailed description of transplant
outcomes for patients in the control group as well as graft survival in all transplant
recipients.
Matched Controls
Patients receiving an HLAi tend to have different characteristics compared to those
receiving other types of treatment. Therefore, we matched each of the HLAi recipients to
several controls using multivariate matching. Matching variables were selected on the basis
of initial univariate analyses, as well as evidence from previous studies(17). Specifically,
6
these were age, gender, cRF score, blood group (type 0 and type A), diabetes as primary
cause of ESRD, length of registration in paired scheme, number of previous transplants and
duration of end stage renal disease (ESRD). We defined the latter as time since first chronic
dialysis or first previous transplantation - whichever came earlier or 0 if the patient had
received neither a transplant nor dialysis treatment at the index date. Since 70% of patients
who received a transplant had the same cRF score at listing as at the day of transplant and
80% had a cRF score within ±10 of their listing score, we assumed that cRF was constant
over time for the purpose of matching.
We performed risk-set matching in an attempt to approximate what would have happened
to patients who were observed to have received an ‘HLAi’ transplant had they not received
this treatment (18).In other words, we determined the time t from the date of registration
on the waiting list that it took for each HLAi recipient to obtain a transplant and
constructed a custom pool of potential matches for each of these index cases. This pool
included patients who had not received another transplant or died within the same time
interval after wait listing but excluded patients for whom time t fell before 1.1.2007
because, by definition, our sample only contained patients who had not received a
transplant or died before this date. In the subsequent analysis we compared outcomes of
HLAi patients after the date of transplant, with that of matched controls from time t after
wait listing.
Based on recommendations by Ho et al., we did not assess the balance between the
treatment and the control group using statistical tests (19). Instead, we compared the mean
values, the distribution of continuous variables using quantile-quantile plots and
differences at the individual level for categorical and binary variables. Further
methodological details including a description of the three types of matching approaches
that we have used to test the robustness of our results can be found in the appendix
methods.
Statistical analysis
We performed two comparisons. In the first comparison we censored survival of the
matched controls at the day of transplant (‘Listed only’). In the second comparison, termed
7
‘Listed or Transplant’, we used the same matched control group but followed them until
31.12.2014 or death.
We used the Kaplan-Meier method, and the log-rank test to compare rates of patient
survival between the treatment group ‘HLAi’ and ‘Listed only’ and ‘Listed or Transplant’ as
well as the rate of graft survival between ‘HLAi’ and those matches who received a
compatible deceased donor (cDD) or a compatible living donor (cLD) at some point. As an
alternative and more readily interpretable summary measure of our primary outcome, we
also calculated the difference in restricted mean patient survival, that is, the difference in
average survival time between the groups at 3, 5 and 7 years (20). In this analysis we also
adjusted for any significant imbalances that remained between the groups after matching,
with ‘significant’ defined as a mean difference of more than one tenth of a standard
deviation. As indicated above, we performed these primary analyses using three different
matching approaches to assess the sensitivity of results. Given the known difficulties of
organ allocation and transplantation of patients with the highest cRFs (>85%) as described
earlier, we also conducted a subgroup analysis of patients with cRF of >85% using our base
case matching approach.
Since it is not possible to receive a cLD transplant in the same registry period after having
received a cDD transplant or vice versa, we used a competing risk framework to estimate
the cumulative incidence of transplant status in the matched controls (21), Figure 2. As
previously noted, membership in the KPD scheme is likely to alter the probability of
receiving a transplant. In addition, some residual imbalance in this respect remained after
matching due to the large difference in paired scheme status between HLAi’s and potential
matches (30% vs. 3% registered). Therefore, we reported both the transplant outcomes for
three categories of paired scheme status, i.e. unregistered matches, matches registered for
<12 months and matches registered for >12 months in the paired scheme at time t as well
as a weighted average of these based on the proportions of HLAi’s in each category at time
t.
We used Stata v14·1 and R v3·2·3 for our analyses.
Results
8
We obtained listings for 25,518 patients. Of the 478 patients registered as receiving HLAi
kidney transplants, 137 were excluded as they were from deceased donors, and 123 did not
have a positive cross match. Five patients were excluded from this group as they were
originally listed on the paediatric waiting list.
Therefore, the treatment group was 213 adult HLAi recipients of cross-match positive LD
kidneys, transplanted in the UK from 1.1.2007 until 31.12.2013. The mean age at the time
of transplant was 44yrs ± 11yrs. The majority were female (65·7%) with a median cRF of
96% (IQR 20). Over half (53·1%) of HLAi recipients had received a previous renal
transplant, 46·9% were blood group O, mean duration of ESRD was 7·9 years (IQR 14·6)
see Table 1.
From the remaining 25,040 listed patients, 801 were excluded as they were originally
listed on the paediatric waiting list; 370 received ABO-incompatible transplants and 16 had
gender data missing. See Fig. 1 for full inclusion and exclusion details.
Matched controls
Using the matching methods described for our base case analysis, we then identified 852
matched controls from the remaining 23,865 patients. The differences in average patient
characteristics between the HLAi treatment group and, the matched controls were
generally lower than one tenth of a standard deviations and the ratio of variances was
between 1 and 1 1 (see Table 1)(22). In addition, as reflected in Figures A1-A8 the
distribution of age, gender, blood group, cRF, time on dialysis, diabetes as cause of ESRD
and previous transplantation was comparable between HLAis and matched controls (see
Appendix). The only exception from this was that approximately 12% more patients in the
‘HLAi’ group were member of the paired scheme at the time of transplant.
Transplant status and UK Living Donor Sharing Scheme membership (UKLDKSS)
In the HLAi -treatment group, 31% had been registered in the UKLDKSS, of these five
patients received a transplant through the UKLDKSS which combined HLAi with KPD. In
the matched controls, 18·7% of patients were registered in the UKLDKSS at the time of
matching and 20·8% were ever registered in the UKLDKSS. Adjusting for imbalances in
paired scheme enrollment by reweighing the sample, and the competing risks of cLD and
9
cDD (see methods), at just under 5 years after the matching date approximately 41% of the
matched controls remained untransplanted, 39% had received a cDD and 20% a cLD
transplant.(Fig 2). Matches registered in the paired scheme for less than 12m received the
highest proportion (43%) of compatible LD (cLD) transplants within the 1st year of the
study; for those who were registered for over 1 year in the UKLDKSS, this was substantially
lower (12%), while for those unregistered in the paired scheme, the rate of cLD transplants
remained low and constant over the study period, Appendix Figure 9. The observed effect
of the success of the UKLDKSS in the first 12m after registration is reflected in the similar
likelihood of a cDD or a cLD in the matched cohort within the first year of the study,
however this is not sustained over time, Fig 2.
Graft survival for HLAi recipients compared to matched controls
Mean death-censored graft survival was consistently highest for cLD recipients, and lowest
for HLAi. At 1 year post transplant, graft survival rates were 88% for HLAi, 96% for cLD
and 92% for cDD. By 5 years graft survival was 68% for HLAi; 89% for cLD and 77% for
cDD. These differences were statistically significant with p=0.002 and p=0.046 for the log-
rank tests of HLAi graft survival compared to cLD and cDD respectively, Fig 3.
Patient survival for HLAi recipients compared to matched controls
Overall, there was no statistically significant difference in patient survival between those
receiving an HLAi transplant, and either ‘Listed or Transplant’ (Log rank, p = 0·984) or
‘Listed only’ (Log rank, p = 0·446). Comparing HLAi to 'Listed or Transplant', restricted
mean survival time was 0·2 months higher in the latter at 3 years (SD 0·4, 95% CI -0·6 to 1,
p=0 649); 0·1 months higher at 5 years (SD 0·8, 95%CI -1·5 to 1·8, p=0·893) and – 0·1
months lower at 7 years (SD 0·9, 95% CI -1·9 to 1·6, p=0·866), Fig 4 & Table 2. The
sensitivity analyses in the appendix and the subgroup analysis restricted only to patients
with a cRF of >85% support these findings, Fig 5 & Table 2, Appendix Figures 9-11.
Discussion
For patients receiving a cross match positive HLAi transplant in the UK, there is no patient
survival benefit compared to patients with the same degree of HLA sensitization who wait
for a compatible alternative organ from either the deceased donor list or through the KPD.
In comparison to the situation reported in the US (14), this equivalence is marked, and
10
appears to be due to the excellent dialysis survival outcomes for these highly sensitised
patients listed for an organ. Two other things are apparent. First, just under half (43%) of
patients who wait for a compatible organ are unlikely to be transplanted under 5 years
years after listing; secondly, our data confirms that immediate short-term graft survival of
HLAi transplants is acceptable, but in the longer term, graft survival is poorer than
compatible alternatives.
These data are suggestive of a more nuanced picture of risk and benefit to be presented by
clinicians to highly sensitised patients with an HLA incompatible living donor. Although an
HLAi incompatible transplant does not confer a survival benefit, it does represent a choice
of a timely transplant for patients who otherwise have a great deal of uncertainty with
respect to their future. In this model, supporting current UK practice, for patients with an
HLAi donor who have not yet registered in the paired scheme, the best of a chance of a
compatible living donor transplant through the UKLDKSS is within 12 months of
registration, while beyond this the rate of allocation plateaus (Appendix Figure 7).
There are some inevitable limitations in a retrospective study based on registry data. We
have defined HLAi as transplants that are cross match positive because of a donor specific
antibody (DSA). The definition of a positive cross match, when performed either by FXCM,
or CDC may vary by technique and local laboratory values. Clearly a centrally standardized
cross match would give further confidence regarding the definition of ‘HLAi’ status, but this
was not possible. Estimating the effect of this on our results is difficult, however, limiting
our analysis to only cross match positive ensures a consistency of immunological risk,
given how good outcomes are for patients who are positive for DSA but have a negative
FXCM (12).
We have used diabetes as the primary cause of ESRD as a matching variable, however it
should be noted that although co-morbidity data relating to diabetes would have been
highly desirable, this data is not routinely collected by NHSBT.
Most transplant units in the UK recommend that patients with an HLAi living donor spend a
period of time in the UKLDKSS before considering desensitization and a direct HLAi
transplant. Data from NHSBT suggests that if a patient has been unsuccessful in obtaining a
match in the UKLDKSS within 4 runs, then success is extremely unlikely (NHSBT data). Our
11
data support this, and indicate that the UKLDKSS, has a limited efficacy for sensitised
patients to achieve a compatible transplant, in keeping with NHSBT data, which suggests
that highly sensitised patients are over represented in the UKLDKSS pool, but under
represented in the patients matched and transplanted through the scheme. Clearly our
study period began at the outset of the UKLDKSS when initial numbers of patients included
in each run was low, and the impact is likely to improve with time. Nevertheless, our data
imply that waiting for a compatible organ might not be in the best interests of all patients; a
prospective study of this group of patients is clearly needed to answer the question of
whether the paired scheme helps in obtaining a compatible transplant.
An important limitation of this and other cohort studies is the possibility of bias, both in
methodology and selection. Given the relatively small survival differences between the
‘Dialysis only’ and ‘Dialysis or transplant’ analysis and the stability of estimates across the
sensitivity analyses, it appears that the results were robust to the post-matching imbalance
in paired scheme status. Even if the matched controls were identical in both their observed
and unobserved characteristics to the HLAi treatment group, these patients may be
different from highly sensitised patients not undergoing treatment. In order to explore the
significance of this bias, information would need to be available on whether or not a wait-
listed patient had a living donor, whether incompatible or not, and the date that the living
donor was deemed suitable for transplantation by the transplanting unit. Such a study
would require prospective observations since this information is not currently part of the
NHSBT database.
Death is a relatively crude, but hard end point, which is commonly used as a
transplantation outcome. For a patient weighing up the risks and benefits of
transplantation, the likelihood of survival and quality of life on or off dialysis, are probably
the most important factors, rather than graft survival. We know that in general, quality of
life improves after transplantation(23), but this topic is currently under prospective
multicentre investigation in the UK(24), and there are good reasons to think that quality of
life measures may be different in highly sensitised patients, in whom such a study has not
been performed.
12
In contrast to our analysis, a recent US multicenter study by Orandi et al suggested a
significant survival benefit for HLAi transplantation compared to remaining on the waiting
list and an even greater benefit compared to remaining on dialysis (14, 17). The
characteristics of our HLAi treatment group appear very similar to the equivalent group in
the US study in terms of age, sex, blood group, number of previous transplants, and
duration of renal replacement. Importantly, the latter was 7·9 years in our cohort and 7·3
in the US cohort. However, a methodological difference is that Orandi et al. constructed a
separate ‘Dialysis only’ matched cohort selecting from the subset of patients who never
received a transplant. We would argue that this approach is likely to have overstated the
benefit of HLAi compared to staying on dialysis because the survival time is not only
dependent on transplant status but also affects the likelihood of receiving a transplant in
the first place as a result of time spent on the waiting list. In other words, a reverse
causality exists between transplant status and survival time. Thus, even if receiving a
transplant did not increase survival, we would expect the ‘Dialysis only’ cohort, as selected
in the aforementioned way, to have inferior survival rates, because, by definition, it
excludes individuals who did not survive long enough on the waiting list to receive a
transplant. In statistical terms, this may be considered to be a variant of immortal time bias
(25, 26). In our analysis, we selected a single matched cohort disregarding subsequent
transplant status. For the comparison of HLAi transplantation versus remaining on dialysis,
we then censored survival of the matched controls at the day of transplant. We believe that
this approach provides a better assessment of survival differences.
Beyond this methodological consideration, we do not believe that the findings by Orandi et
al. can be extrapolated outwith the USA. In fact, the results may be more a reflection of the
quality of dialysis care in the USA than of a distinct advantage of HLAi as a transplant
modality (15). The reason for this difference is not easily explained, but as Foley et al refer
to, they are certainly multifactorial and complex – involving the differences beyond the
technical feats of dialysis provision, and pertaining to the differences in holistic
management of the dialysis patient afforded within different health systems(16).
To this end, our study provides a more circumspect view of the benefit of undertaking HLA-
incompatible transplantation. This transplant modality is not to be undertaken lightly,
since, as we show, there is a higher risk of graft loss, compared to compatible
13
transplantation undertaken in similar patients (Figure 3). Additionally, our data indicate
that, for patients who have not been registered in the paired scheme, at the time of
considering an HLAi transplant, the chance of a compatible living donor transplant from the
paired scheme are best within the 1st year of registration in the UKLDKSS (Appendix Figure
A9). Otherwise, the overall rate of compatible transplantation, whether cDD or cLD,
remains constant over time (Figure 2).
Sensitised patients comprise a significant proportion of the waiting list, and attempts to
improve the chances of deceased donor organ allocation are important steps in achieving
equity of access(27). For some highly sensitised patients, the finding demonstrated by
these data that they are likely to wait for 4 years on dialysis before half of the cohort
receive a compatible transplant may be sufficiently unacceptable to make them willing to
take the risk of an HLAi, while for others this may not be the case. As clinicians, our task is
to inform that choice and further prospective work to understand patient motivations and
concerns, and quality of life issues in this difficult-to-transplant population is needed.
In summary, this is the first European national study to examine the question of survival
after HLAi transplantation, and specifically address the issue of whether highly
immunologically sensitised patients are best served by accepting a living donor HLAi after
desensitization or waiting for a compatible alternative. For patients who opt to wait, the
most likely outcome is to remain untransplanted for at least 4 years. For those who do
choose to go ahead with desensitization and HLAi transplant, their survival is comparable
to those who wait, but there is no benefit.
Acknowledgements:
The authors would like to acknowledge the support of NHSBT, as well as the work of all UK
Transplant units contributing to this database.
The research was funded/supported by the National Institute for Health Research (NIHR)
Biomedical Research Centre based at Guy's and St Thomas' NHS Foundation Trust and
King's College London. The views expressed are those of the author(s) and not necessarily
those of the NHS, the NIHR or the Department of Health.
Additional thanks to Jean Kwun for critical manuscript review.
14
Research in context
Evidence before this study
We searched PubMed and Medline, links from published papers and authors’ personal
libraries for studies of the effect of HLAi kidney transplantation on patient survival. The
terms searched for in PubMed were: ‘kidney transplantation AND HLA-incompatible
transplantation AND survival’, ‘kidney transplantation AND positive cross match AND
survival’. Of the 105 studies cited, publications relating to single cases or centres, as well as
combining ABO-incompatible outcomes were excluded. One study was found.
Added value of this study
This study is the first European study to provide robust evidence for survival outcomes for
sensitised patients awaiting a transplant.
Implications of all the available evidence
HLA-incompatible kidney transplantation in the UK confers no greater risk to patient
survival than awaiting a compatible living or deceased donor alternative, but does not
confer a survival benefit either. For sensitised patients with an incompatible (i.e.
crossmatch positive) living donor, it represents the most guaranteed route to
transplantation. While clinicians have focussed on graft and patient survival as measures of
success in transplantation, more study of patient quality of life measures related to dialysis
or transplantation are needed to evaluate the risk-benefit of HLA-incompatible
transplantation in highly sensitised patients who are otherwise likely to remain
untransplanted.
15
Tables
Table 1. Demographics of ‘HLAi’ treatment group compared to matched controls –
matching criteria (Base case analysis)
Characteristic ‘HLAi’n = 213
Matched Controlsn = 852
Difference in means (Ratio of variances)
Age (years)(mean ± SD)
44 (±11) 46 (±12) -1·7 (1)
cRF (%)
median (25th – 75th
percentile)
0 – 20%, as % of group21 – 50%, as % of group 51 – 80%, as % of group80 – 100%, as % of group
96 (79 – 99)
8·94·713·173·2
91 (71 – 97)
8·76·714·370·3
3·4 (1)
0·2-2-1·22·9
Duration of ESRD (years)median (25th – 75th percentile)
7·9 (1·3 – 15·9) 8 (0·9 -15·3) 0 (1)
No of prev tx (% of patients)0123
46·93911·72·3
48·437·912·11 6
-1·51·1-0·40·7
Female (% of patients) 65·7 65·6 0·1
Blood Group(% of patients)
OA
ABB
46·933·315·54·2
47·833·515·53·3
-0·9-0·200·9
Diabetes as primary cause of ESRD (% of patients)
0·5 0·5 0
Length of timeregistered in pairedscheme
0 (not registered)0 – 12 months
>12 months
69·513·616·9
81·38·99·7
-11·84·77·2
16
Table 2. Adjusted absolute and differences in restricted mean patient survival of HLAi compared to matched cohort ‘Listed only’ and ‘Listed or transplant’ in months
Scenario (nHLAi/nmatches ¿ Years
HLAi Listed or transplant
Listed only HLAis relative to ‘Listed or transplant’
HLAis relative to ‘Listed only’
Mean (SE)
Mean(SE)
Mean (SE) Mean (SE)
95%CI Mean(SE)
95%CI
Base
case
ana
lysis
(2
13/8
52)
3 34·4(0·4)
34·7(0·2)
34·5(0·2)
0 2(0 5)
-0·8 to1·2
0(0·5)
-0·9 to0·9
5 56·7(0·8)
56·8(0·4)
56 2(0 5)
0 1(0 8)
-1·5 to1·8
-0·5(0·9)
-2·3 to1·3
7 78·3(1·3)
78 3(0·7)
76·9(1)
-0·1(1·9)
-3·8 to3 5
-1 5(1 9)
-5·3 to2 3
10 to
1 m
atch
ing
(213
/213
0)
3 34·2(0·5)
34(0·2)
33·9(0·2)
-0·2(0·4)
-1 to0·6
-0 3(0 5)
-1 4 to0 7
5 56·1(0·9)
55·3(0 3)
54·6(0 4)
-0·9(1·1)
-3·1 to1·3
-1·4(1)
-3·4 to0·6
7 77·7(1 4)
75·7(0 6)
74·3(0·7)
-2(1·9)
-5·7 to1·7
-3(1·7)
-6·4 to0·3
Coar
sed
exac
t mat
chin
g (1
43/4
718)
3 34 8(0 4)
34·7(0 3)
34·6(0·3)
-0·1(0·5)
-1·1 to0·9
-0·2(0·5)
-1·1 to0·7
5 57(0·9)
56·9(0·6)
56·5(0·7)
-0·1(1·4)
-2·8 to2 6
-0 5(1 1)
-2·6 to1·6
7 78·6(1 5)
78·5(1)
77·6(1·3)
-0·1(2·4)
-4·8 to4·7
-0·9(2·4)
-5·6 to3·8
>85%
cRF
scor
e on
ly
(152
/608
)
3 33·9(0·6)
34·2(0·3)
34(0 3)
0·2(0·7)
-0 7 to1 7
0(0·7)
-1·3 to1·4
5 55·9(1·1)
55·8(0 6)
55 2(0 8)
-0·2(1·6)
-3·2 to2·9
-0·7(1·4)
-3·5 to2
7 77·8(1·8)
76·8(1·1)
75·7(1·4)
-1(2·2)
-5·2 to3·3
-2·1(2)
-6 to1·8
17
References
1. Pruthi R, Hilton R, Pankhurst L, Mamode N, Hudson A, Roderick P, et al. UK Renal Registry 16th annual report: chapter 4 demography of patients waitlisted for renal transplantation in the UK: national and centre-specific analyses. Nephron Clinical practice. 2013;125(1-4):81-98.2. Patel R, Terasaki PI. Significance of the Positive Crossmatch Test in Kidney Transplantation. N Engl J Med. 1969;280(14):735-9.3. Johnson RJ, Allen JE, Fuggle SV, Bradley JA, Rudge C, Kidney Advisory Group UKTN. Early experience of paired living kidney donation in the United kingdom. Transplantation. 2008;86(12):1672-7.4. Sharif A, Kraus ES, Zachary AA, Lonze BE, Nazarian SM, Segev DL, et al. Histologic phenotype on 1-year posttransplantation biopsy and allograft survival in HLA-incompatible kidney transplants. Transplantation. 2014;97(5):541-7.5. Bentall A, Cornell LD, Gloor JM, Park WD, Gandhi MJ, Winters JL, et al. Five-year outcomes in living donor kidney transplants with a positive crossmatch. Am J Transplant. 2013;13(1):76-85.6. Iyer HS, Jackson AM, Zachary AA, Montgomery RA. Transplanting the highly sensitised patient: trials and tribulations. Curr Opin Nephrol Hypertens. 2013;22(6):681-8.7. Warren DS, Montgomery RA. Incompatible kidney transplantation: lessons from a decade of desensitization and paired kidney exchange. Immunol Res. 2010;47(1-3):257-64.8. Sharif A, Zachary AA, Hiller J, Segev D, Alachkar N, Kraus ES, et al. Rescue kidney paired donation as emergency salvage for failed desensitization. Transplantation. 2012;93(7):e27-9.9. Montgomery RA, Lonze BE, Jackson AM. Using donor exchange paradigms with desensitization to enhance transplant rates among highly sensitised patients. Curr Opin Organ Transplant. 2011;16(4):439-43.10. Susal C, Morath C. Current approaches to the management of highly sensitised kidney transplant patients. Tissue Antigens. 2011;77(3):177-86.11. Jackson AM, Leffell MS, Montgomery RA, Zachary AA. A GPS for finding the route to transplantation for the sensitised patient. Curr Opin Organ Transplant. 2012;17(4):433-9.12. Orandi BJ, Garonzik-Wang JM, Massie AB, Zachary AA, Montgomery JR, Van Arendonk KJ, et al. Quantifying the risk of incompatible kidney transplantation: a multicenter study. Am J Transplant. 2014;14(7):1573-80.13. Vo AA, Petrozzino J, Yeung K, Sinha A, Kahwaji J, Peng A, et al. Efficacy, outcomes, and cost-effectiveness of desensitization using IVIG and rituximab. Transplantation. 2013;95(6):852-8.14. Orandi BJ, Luo X, Massie AB, Garonzik-Wang JM, Lonze BE, Ahmed R, et al. Survival Benefit with Kidney Transplants from HLA-Incompatible Live Donors. N Engl J Med. 2016;374(10):940-50.15. Goodkin DA, Mapes DL, Held PJ. The Dialysis Outcomes and Practice Patterns Study (DOPPS): How Can We Improve the Care of Hemodialysis Patients? Semin Dial. 2001;14(3):157-9.16. Foley RN, Hakim RM. Why is the mortality of dialysis patients in the United States much higher than the rest of the world? J Am Soc Nephrol. 2009;20(7):1432-5.17. Montgomery RA, Lonze BE, King KE, Kraus ES, Kucirka LM, Locke JE, et al. Desensitization in HLA-Incompatible Kidney Recipients and Survival. N Engl J Med. 2011;365(4):318-26.18. Rosenbaum PR. Risk-Set Matching. Design of Observational Studies. New York, NY: Springer New York; 2010. p. 223-35.19. Ho DE, Imai K, King G, Stuart EA. Matching as Nonparametric Preprocessing for Reducing Model Dependence in Parametric Causal Inference. Political Analysis. 2007;15(3):199-236.20. Royston P, Parmar MK. Restricted mean survival time: an alternative to the hazard ratio for the design and analysis of randomized trials with a time-to-event outcome. BMC medical research methodology. 2013;13:152.21. Hinchliffe SR, Lambert PC. Extending the flexible parametric survival model for competing risks. Stata J. 2012;12:674-87.22. Stuart EA, and Donald B. Rubin. . Best practices in quasi-experimental designs. In: Osborne JW, editor. Best Practices in Quantitative Methods2008. p. 155-76.23. von der Lippe N, Waldum B, Brekke FB, Amro AA, Reisaeter AV, Os I. From dialysis to transplantation: a 5-year longitudinal study on self-reported quality of life. BMC nephrology. 2014;15:191.
18
24. Oniscu GC, Ravanan R, Wu D, Gibbons A, Li B, Tomson C, et al. Access to Transplantation and Transplant Outcome Measures (ATTOM): study protocol of a UK wide, in-depth, prospective cohort analysis. BMJ open. 2016;6(2):e010377.25. Lévesque LE, Hanley JA, Kezouh A, Suissa S. Problem of immortal time bias in cohort studies: example using statins for preventing progression of diabetes2010 2010-03-12 12:16:39.26. Kim SJ. Immortal Time Bias in Cohort Studies of Kidney Transplant Recipients. Am J Transplant. 2010;10(1):190-.27. Nguyen HD, Wong G, Howard K, Claas FH, Craig JC, Fidler S, et al. Modeling the benefits and costs of integrating an acceptable HLA mismatch allocation model for highly sensitised patients. Transplantation. 2014;97(7):769-74.
19
Appendix Methods & TablesMatching methods involve an implicit trade off between bias and variance and no
approach can be proven to be ‘optimal’ in any specific sample. In our base case analysis
we used mahalanobis distance (MD) matching with a target matching ratio 4 to 1 (1).
Firstly, we simultaneously calculated the MD of the HLAIs with respect to the potential
matches within their respective risk-matching sets. Secondly, we determined the pairing
of treated and matched controls that minimized the total sum of mahalanobis distance
and matched each control to at most one treated unit (2). If there were multiple matches
with the same MD as the largest MD among those selected, we included all of these
observations using sampling weights in the analysis to still obtain the same sum of
sample weights for each HLAi. As an alternative to our base case analysis, we matched
each HLAi recipient with ten rather than four controls using the same MD matching
approach thus effectively relaxing our matching criteria (Table 2). In addition, rather
than matching based on estimated statistical distance between observations we
matched on criteria defined by us using Coarsened Exact Matching (CEM) with a variable
matching ratio (3). In this sensitivity analysis, we effectively matched on stricter criteria
than in the base case analysis but only on a subset of the HLAi cohort (approximately
2/3 of the HLAi sample) for whom such closer matches were available. More specifically,
we divided non-binary matching variables into groups. For age these were 18-29, 30-39,
40-49, 50-59, 60-69 and 70-85 years, for duration of ESRD 0-5, 6-11, 12-23, 24-47, 48-
71 and 72+ months, for duration of listing on paired scheme, 0, 0-12, 12+ months and
for the cRF score 0, 1-59, 60-84, 85- 94 and 95-100. In addition, we matched on whether
patients had a previous transplant, whether diabetes was the primary cause of ESRD,
and matches we required to have the same gender, the same blood type (0, A or other)
in this sensitivity analysis. Subject to at least one match remaining for each HLAi, any
control that was matched to multiple treated patients was assigned to the one with the
lowest MD. Again, we used sampling weights to reweigh the sample depending on the
number of matches per HLAi recipient (Table 3)
26
Appendix Table 1. Mean differences of subgroup analysis of patients with a cRF of >85%
Characteristic ‘HLAi’n = 152
Matched Controlsn = 608
Difference in means (Ratio of variances)
Age (years)(mean ± SD)
43 (±11) 45 (±11) -2·2 (0·9)
cRF (%)
median (25th – 75th
percentile)
0 – 20%, as % of group21 – 50%, as % of group51 – 80%, as % of group80 – 100%, as % of group
98 (95 – 100)
000100
97 (94 – 99)
000100
0·9 (0·9)
0000
Duration of ESRD (years)median (25th – 75th percentile)
11·1 (3·1 -18·1) 10·9 (3·8 -18·5) -0·2 (0·9)
No of prev tx (% of patients)0123
34·246·715·83·3
3643·318·12·6
-1·8 3·4-2·3 0·7
Female (% of patients) 62·5 60·4 2·1
Blood Group(% of patients)
OA
ABB
48·732·913·84·6
46·934·014·14·9
1·8-1·1-0·3-0·3
Diabetes as primary cause of ESRD (% of patients)
0·7 0·7 0
Length of timeregistered in pairedscheme
0 (not registered)0 – 12 months
>12 months
62·515·821·7
84·07·48·6
-21·5 8·4 13·1
27
Appendix Table 2: Mean differences of matching run using 10 to 1 matches to index cases
Characteristic ‘HLAi’n = 213
Matched Controlsn = 2130
Difference in means (Ratio of variances)
Age (years)(mean ± SD)
44 ± 11 48 ± 12 -3·1 (1·)
cRF
median (25th – 75th percentile)
0 – 20%, as % of group21 – 50%, as % of group51 – 80%, as % of group80 – 100%, as % of group
96 (79 - 99)
8·94·713·173·2
87 (57 -96)
9·710·916·662·8
7·9 (0 9)
-0·8-6·2-3·510·4
Duration of ESRD (years)median (25th – 75th percentile)
7·9 (1·3 – 15·9) 7 (0·8 – 14·9) 0·3 (1)
No of prev tx (% of patients)0123
46·939·011·72·3
51·335·911·31·5
-4·43·10·40·8
Female (% of patients) 65·7 64·7 1
Blood Group(% of patients)
OA
ABB
46·933·315·54·2
45·834·615·73·9
1·1-1·3-0·2 0·3
Diabetes as primary cause of ESRD (% of patients)
0·5 0·5 0
Length of timeregistered in pairedscheme
0 (not registered)0 – 12 months
>12 months
69·513·616·9
88·27·24·6
-18·76·412·3
28
Appendix Table 3: Mean differences using Coarsened Exact Matching (CEM) methods
Characteristic ‘HLAi’n = 143
Matched Controlsn = 4718
Difference in means (Ratio of variances)
Age (years)(mean ± SD)
46 (± 12) 46 (± 12) -0 7 (1)
cRF
median (25th – 75th
percentile)
0 – 20%, as % of group21 – 50%, as % of group51 – 80%, as % of group80 – 100%, as % of group
92 (60 – 99)
12·67·017 562 9
91 (61 – 98)
15·36·714·963·1
1·6 (0·9)
-2·7 0·3 2·6-0·2
Duration of ESRD (years)median (25th – 75th percentile)
0·4 (0·1 – 1·2) 0·4 (0 – 1·2) 0 (0·9)
No of prev tx (% of patients)0123
54·530·811·92·8
54 53013·42
0 0·8-1·5 0·8
Female (% of patients) 67·8 67·8 0
Blood Group(% of patients)
OA
ABB
46·936·412·64·2
46·935·912·64·7
00·50-0·5
Diabetes as primary cause of ESRD (% of patients)
0 0 0
Length of timeregistered in pairedscheme
0 (not registered)0 – 12 months
>12 months
89·55·64·9
89·55·64·9
000
29
References
1. Kantor D. MAHAPICK: Stata module to select matching observations based on a Mahalanobis distance measure. Statistical Software Components2012/11/15/.2. Hansen BB, Fredrickson, M., Fredrickson, M.M.M., Rcpp, L. and Rcpp, I. Package ‘optmatch’. 2016.3. Blackwell M IS, King G, Porro G. Cem: Coarsened exact matching in stata. Stata J 2009;9:524–46. Epub 524.
30
Figure A1. Quantile-quantile plot for calculated reaction frequency (cRF) between HLAi cohort and identified matches (N.B.: for clarity, observation have been randomly perturbed by a small amount)
2040
6080
Age
in y
ears
(HLA
is)
20 40 60 80Age in years (Matched controls)
Figure A2. Quantile-quantile plot showing difference in age between the HLAi cohort and identified matches (N.B.: for clarity, observations have been randomly perturbed by a small amount)
020
4060
8010
0C
RF
scor
e (H
LAis
)
0 20 40 60 80 100CRF score (Matched controls)
31
45.7 1.3
1.8 28.8 0.1 2.7
0.1 15.4
0.2 3.4 0.6
O
A
B
AB
Blo
od g
roup
(HLA
Is)
O A B ABBlood group (Matched controls)
Figure A3. Difference in blood group between HLAi cohort and identified matches
33.3 0.9
1.1 64.7
Female
Male
Gen
der (
HLA
is)
Male FemaleGender (Matched controls)
Figure A4. Difference in gender between HLAi cohort and identified matches
32
44.2 2.6 0.1
4.0 33.8 1.1 0.1
0.1 1.5 10.1
0.8 1.53
2
1
0
Num
ber o
f pre
viou
s tra
nspl
ants
(HLA
is)
0 1 2 3Number of previous treatments (Matched controls)
Figure A5. Difference in number of previous transplants between HLAi and identified matches
010
2030
40D
urat
ion
of E
SR
D in
yea
rs (H
LAis
)
0 10 20 30 40Duration of ESRD in years (Matched controls)
Figure A6. Quantile-quantile plot for time on dialysis between HLAi cohort and identified matches (N.B.: for clarity, observations have been randomly perturbed by a small amount)
33
99.5
0.51
0
Dia
bete
s re
cord
ed a
s pr
imar
yca
use
of E
SR
D (H
LAis
)
0 1Diabetes recorded as primary
cause of ESRD (Matched controls)
Figure A7. Difference in diabetes as primary cause of ESRD between HLAi and identified matches
69.5
4.7 8.9
7.2 9.7
0
0-12
13+Tim
e re
gist
ered
in p
aire
d sc
hem
e in
mon
ths
(HLA
is)
0 0-12 13+Time registered in paired scheme in months (Matched controls)
Figure A8. Differences in time registered in paired scheme between HLAis and matched controls
34
020406080
100
% o
f mat
ched
con
trols
0 12 24 36 48 60 72 84 96Months
Registered <12 monthson the date of matching
020406080
100
% o
f mat
ched
con
trols
0 12 24 36 48 60 72 84 96Months
Registered >12 monthson the date of matching
020406080
100
% o
f mat
ched
con
trols
0 12 24 36 48 60 72 84 96Months
Not registered in the pairedscheme on the date of matching
Transplanted with an LD compatibleTransplanted with a DD compatible
Figure A9. Figure 2. Matched cohort: expected transplant status of matched controls during the study period, with 95% confidence bands. A. transplant status for patients registered in the paired scheme (’UKLDKSS) for less than 1 year on the date of matching. B. transplant status for patients registered for over 1 year in the UKLDKSS for over 1 year on the date of matching in the study. C.
BA
C
35
transplant status for patients not registered in the paired scheme on the date of matching
0
20
40
60
80
100
Pat
ient
sur
viva
l (%
)
2130 1243 755 436 245 141 74 33 0Dialysis only2130 1631 1234 921 674 483 299 155 0Dialysis or transplant213 202 178 147 115 88 48 19 0HLAi transplant
Number at risk
0 12 24 36 48 60 72 84 96Months
HLAi transplant Dialysis or transplantDialysis only
Figure A10. Kaplan-Meyer survival curve comparing HLAi against a matched cohort using 10 to 1 mahalanobis distance matching
0
20
40
60
80
100
Pat
ient
sur
viva
l (%
)
4718 3262 2173 1442 866 527 283 129 0Dialysis only4718 4002 3225 2488 1893 1363 876 414 0Dialysis or transplant143 138 127 103 83 64 38 14 0HLAi transplant
Number at risk
0 12 24 36 48 60 72 84 96Months
HLAi transplant Dialysis or transplantDialysis only
Figure A11. Kaplan-Meyer survival curve comparing HLAi against a matched cohort using coarse exact matching
36
0
20
40
60
80
100
Gra
ft su
rviv
al (%
)
125 86 55 39 27 14DD compatible47 30 19 14 11 6LD compatible151 123 108 87 64 45HLAi transplant
Number at risk
0 12 24 36 48 60Months after transplant
HLAi transplant LD compatibleDD compatible
Figure A12. Comparison of death-censored graft survival in HLAi against a matched cohort of patients receiving a compatible kidney transplant from either a living (cLD) or deceased donor (cDD), subgroup analysis of patients with a cRF of >85%
37