Biomarkers in transplantation: Prospective ... - Kidney International

original article

http://www.kidney-international.org & 2006 International Society of Nephrology

Biomarkers in transplantation: Prospective, blinded measurement of predictive value for the flow cytometry crossmatch after negative antiglobulin crossmatch in kidney transplantation R Wen1,4, V Wu1, S Dmitrienko1, A Yu1, R Balshaw2, PA Keown1,3 and the Genome Canada Biomarkers in Transplantation Group 1

Immunology Laboratory, Vancouver Hospital, Vancouver, Canada; 2Department of Statistics, Simon Fraser University, Vancouver, Canada and 3Division of Nephrology, Department of Medicine, University of British Columbia, Vancouver, Canada

This prospective, blinded observational study was conducted to measure the predictive value the of flow cytometric crossmatch for biopsy-proven acute rejection, graft loss, or death following kidney transplantation. Patients were selected for renal transplantation on the basis of a conventional antihuman globulin cytotoxic T-cell crossmatch. Flow crossmatch was performed simultaneously, but the results were not disclosed to the transplant team. A total of 257 kidney transplant recipients were enrolled in the study; 78 patients experienced biopsy-proven rejection in the first posttransplant year, and 41 patients lost their graft or died during the period of follow-up (mean: 2046 days). Kaplan–Meier estimates of rejection, graft loss, or patient death did not differ between subjects with a positive or negative flow crossmatch. Cox analyses showed no influence of the flow crossmatch on the risk of biopsy-proven acute rejection (P ¼ 0.987). The sensitivity and specificity of the flow crossmatch for prediction of biopsy-proven rejection were 0.128 and 0.883, and the positive and negative post-test probabilities were 0.323 and 0.301, respectively. The magnitude of the channel shift did not influence the multivariate Cox regression model. The area under the receiver operating characteristic curve of the flow crossmatch was 0.483 (P ¼ 0.71) and 0.572 (P ¼ 0.38), respectively for the living and cadaver transplant recipients, indicating no discriminative value in this study population. Flow crossmatch appears to have no significant incremental value in predicting biopsy-proven acute rejection, graft loss, or death following kidney transplantation in patients who have a negative antihuman globulin cytotoxic T-cell crossmatch against their donor. Correspondence: PA Keown, Immunology Laboratory, Vancouver General Hospital, 855, West 12th Avenue, Vancouver BC, V5Z 1M9, British Columbia, Canada. E-mail: [email protected] 4 Current address: Division of Nephrology, the Second Hospital, Shandong University, Shandong, PR China

Received 19 January 2006; revised 1 July 2006; accepted 11 July 2006; published online 30 August 2006 1474

Kidney International (2006) 70, 1474–1481. doi:10.1038/sj.ki.5001785; published online 30 August 2006 KEYWORDS: kidney transplantation; rejection; flow cytometry crossmatch

Sensitive and specific biomarkers are critical in all areas of medical practice to define risk, predict outcome, guide therapy, and improve selection for clinical research.1–3 This is particularly true in renal transplantation, where rapid advances in immunological understanding and intervention are driving important improvements in clinical outcomes. The pre-transplant crossmatch is designed to detect preformed immunoglobulin (IgG) antibodies against donor human lymphocyte antigens (HLAs) in the recipient’s serum which may cause antibody-mediated rejection and may jeopardize graft survival.4–10 Anti-human globulin (AHG) is currently routinely employed to enhance the sensitivity of the conventional complement-dependent cytotoxic (CDC) crossmatch, and is the standard crossmatch method in the majority of laboratories worldwide. The flow cytometry crossmatch (FCXM) was first used for pre-transplant crossmatch in 1983.11 Although the assay offers no direct evidence of biological effect or antigenic specificity, it is generally accepted to be more sensitive than AHG-XM for detecting low levels of anti-lymphocyte antibodies.10–15 On this basis, the FCXM is now widely employed as a key pre-transplant measure, and a positive FCXM has been proposed as a valid rationale to withhold transplantation, to augment immunosuppression, or to employ innovative strategies to reduce sensitization.16–18 A growing body of studies has examined the association between FCXM and acute rejection in the last decade. As reviewed in detail by Gebel and Bray,16 many of these reports have suggested that FCXM-positive recipients experienced a greater incidence of acute rejection and early graft loss, especially among the subgroups of recipients at high immunological risk,19–23 whereas other studies have shown negative results or indicated little advantage of FCXM over Kidney International (2006) 70, 1474–1481

original article

R Wen et al.: Flow crossmatch in renal transplantation

the AHG-XM in terms of reducing acute rejection episodes and graft loss.24–27 Few studies, however, have attempted to define the predictive value of this assay in objective terms,28,29 and none has undertaken this in a prospective and blinded manner. The aim of this study was therefore to measure the predictive value of donor T cell-directed IgG FCXM for acute rejection and graft loss following kidney transplantation under conditions of normal clinical practice. The study was conducted in both living donor (LD) and cadaveric donor (CD) recipients. FCXM was performed before transplant but the result remained confidential and was not employed for clinical decision purposes. RESULTS Patients

Patient characteristics are shown in Table 1. Of the 257 patients included in the study, 61% (156 patients) were male. Their original diseases were glomerulonephritis (40.1%), diabetic nephropathy (11.3%), cystic kidney disease (9.7%), pyelonephritis or interstitial nephritis (9.3%), lupus nephritis (6.2%), congenital disease (3.1%), and others (20.2%). One hundred and ninety-seven patients (77%) received a first graft and 60 (23%) a second or subsequent graft. Twenty patients received induction therapy with depleting antibody, and 27 (11%) experienced delayed graft function (DGF) requiring dialysis. Compared with living donor transplants, the cadaver donor recipients had a higher panel reactive antibody (PRA) (median: 33 vs 0%), and a greater number of second or subsequent grafts (56 vs 14%). A higher proportion also received biological induction therapy with OKT3 (32 vs 1%). The mean duration of follow-up for the study was 2046 days (range: 0–3112 days). FCXM results

Two hundred and twenty-six patients had a negative T-cell FCXM. Among these, 173 patients had a 0 MCS; 37 had a MCS between one and five MCS, and 16 were between six

Table 1 | Demographic variables of study subjects according to donor source Living donor (N=200) FCXM (median shift) PRA (%) HLA-A,B mismatch number HLA-DR mismatch number Recipient age (years) Donor age (years) Regraft (%) DGF (%) Induction therapy (%) Recipient serum CMV+ (%) Donor serum CMV+ (%)

0 0 2 1 45 42 28 18 2 112 106

(0,2) (0,3) (1,3) (0,1) (34,54) (35,49) (14%) (9%) (1%) (56%) (53%)

Cadaver donor (N=57) 0 33 2 1 38 31 32 9 18 41 24

(0,12) (0,73) (2,3) (1,2) (32,52) (16,51) (56%) (16%) (32%) (72%) (42%)

CMV, cytomegalovirus; DGF, delayed graft function; FCXM, flow cytometry crossmatch; HLA, human lymphocyte antigen; PRA, panel reactive antibody. The values for FCXM, PRA, HLA-A, B mismatch number, HLA-DR mismatch number, recipient age, and donor age were expressed as: median (first quartile, third quartile).

Kidney International (2006) 70, 1474–1481

and nine. A total of 31 (12.1%) patients had a positive T-cell FCXM (range: 10–87 MCS). For two patients who experienced primary graft non-function, the FCXM results were 0 and 18 MCS; for two with graft loss owing to venous thrombosis, FCXM results were 0 and 2 MCS. Several key risk factors were more common in patients with a positive FCXM (Table 2), including PRA levels (median: 63 vs 0%), re-transplantation (42 vs 21%), and cadaver donor (52 vs 18%). Acute rejection

Of the 257 recipients who underwent transplantation between January 1997 and December 2000, a total of 78 (30.4%) developed biopsy proven acute rejection (BPAR) within the first year post-transplant (Table 3). Kaplan–Meier estimates of the cumulative probabilities of BPAR are shown in Figure 1. Of the 31 patients with a positive T-cell FCXM, 33.3% (Kaplan–Meier estimate) had BPAR within the first year compared with 30.7% of the 226 subjects with a negative FCXM (P ¼ 0.907) (Figure 1a). The proportion of patients with BPAR among the recipients of first or subsequent grafts was 29.7 and 35.4% (P ¼ 0.435) respectively. The incidence of

Table 2 | Demographic characteristics of study subjects according to flow crossmatch status Negative FCXM (N=226) Panel reactive antibody (%) HLA-A,B mismatch number HLA-DR mismatch number Recipient age (years) Donor age (years) Retransplant (%) Cadaver donor (%) Delayed graft function (%) Recipient serum CMV+ (%) Donor serum CMV+ (%)

0 2 1 43 42 47 41 23 134 117

Positive FCXM (N=31)

(0,7) (1,3) (0,1) (33,53) (33,49) (21%) (18%) (10%) (59%) (52%)

63 2 1 37 37 13 16 4 19 13

(0,87) (2,3) (1,1) (30,53) (25,54) (42%) (52%) (13%) (61%) (42%)

CMV, cytomegalovirus; FCXM, flow cytometry crossmatch; HLA, human lymphocyte antigen; PRA, panel reactive antibody. The value for PRA, HLA-A, B mismatch number, HLA-DR mismatch number, recipient age, and donor age were expressed as: median (first quartile, third quartile).

Table 3 | Contingency table showing frequency of BPAR and graft failure or death (in brackets) according to patient group and FCXM positivity Patient category Living donor FCXM (+) FCXM () Cadaver donor FCXM (+) FCXM () All patients FCXM (+) FCXM () Total

Event (+)

Event ()

Total

3 (1) 55 (24)

12 (14) 130 (161)

15 185

7 (5) 13 (11)

9 (11) 28 (30)

16 41

10 (6) 68 (35)

21 (25) 158 (191)

31 226

78 (41)

179 (216)

257

BPAR, biopsy proven acute rejection; FCXM, flow cytometry crossmatch.

1475

original article


a positive (46.7%) or a negative (32.6%) FCXM (P ¼ 0.501) (Figure 1c). A total of 111 patients were treated for rejection within the first year post-transplant. Comparable multivariate analyses showed no relationship between the probability of treated acute rejection and FCXM status in any of the strata examined above.

a 1.0 0.9

Prob first event

0.8 0.7 0.6 0.5 0.4 0.3

Graft failure or death

0.2

A total of 41 patients (15.9%) lost their graft (n ¼ 23, 8.9%) or died (n ¼ 18, 7.0%) during the period of follow-up (Table 3). The proportions of patients surviving with a functioning graft are shown in Figure 2. Of the 31 patients with a positive T-cell FCXM, 6 (19.4%) lost their graft or died compared with 35 of the 226 (17.8%) with a negative FCXM, but there was no difference between the two groups at 3000 days (Figure 2a) (P ¼ 0.607). The proportion of patients with graft failure or death among recipients of a first or subsequent graft was 14.8 and 28.5%, respectively (P ¼ 0.003). There was no significant difference in either group between patients with a positive or negative FCXM (Figure 2b) (P ¼ 0.872 for patients with first transplant and P ¼ 0.795 for patients with 41 transplants). The proportion of patients with graft failure or death in the living donor and cadaver donor transplant groups was 15.1 and 28.2%, respectively (P ¼ 0.004). The incidence of graft failure or death was lower in living donor recipients with a positive FCXM (6.7%) than in those with a negative FCXM (15.9%) by 3000 days, but this difference was not statistically significant (P ¼ 0.444). This relationship was reversed in cadaver graft recipients, where the proportion of patients with graft failure or death in patients with a positive FCXM (31.3%) was marginally but not significantly higher than in those with a negative FCXM (26.9%) (P ¼ 0.632). There was no relationship between FCXM and either graft failure or death when these outcomes were analyzed separately.

0.1

0

30 60

90 120 150 180 210 240 270 300 330 360 390 Survival time (days) Negative FCXM

Positive FCXM

b 1.0 0.9 Prob first event

0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

30 60

90 120 150 180 210 240 270 300 330 360 390 Survival time (days)

Positive FCXM and one tx

Positive FCXM and > one tx

Negative FCXM and one tx

Negative FCXM and > one tx

c 1.0 0.9 Prob first event

0.8 0.7 0.6 0.5 0.4 0.3 0.2

Cox regression

0.1 0

30

60

90 120 150 180 210 240 270 300 330 360 390 Survival time (days)

Positive FCXM and CAD

Positive FCXM and LD

Negative FCXM and CAD

Negative FCXM and LD

Figure 1 | Kaplan–Meier estimate of time to first bBPAR by (a) FCXM status, (b) FCXM status and graft number, and (c) FCXM status and donor source.

BPAR was not significantly different between first graft recipients with a positive (27.8%) or a negative (30.0%) FCXM (P ¼ 0.790), or between multiple graft recipients with a positive (41.7%) or a negative FCXM (33.6%) (P ¼ 0.754) (Figure 1b). The proportion of patients with BPAR in the living donor and cadaver donor transplant groups was 29.5 and 36.5%, respectively (P ¼ 0.291). The incidence of BPAR was not significantly different between live donor graft recipients with a positive (20.0%) or a negative (30.2%) FCXM (P ¼ 0.386), or between cadaver graft recipients with 1476

Two patients who did not have complete data on all parameters were excluded from the Cox analysis. Patients with graft failure were censored at the time of graft loss, and those who died with functioning graft were censored at the time of death. The multivariate Cox regression models included T-cell FCXM, PRA level (PRA), HLA-A, B mismatch number, HLA-DR mismatch number, number of prior transplants, immunosuppression, plus donor and recipient ages, and cytomegalovirus serum statuses. The only demographic or treatment-related factor that was a significant predictor for biopsy-proven acute rejection was HLA-A/ B mismatch (hazard ratios (HR) 1.292: P ¼ 0.041). Transplant number was a significant predictor of graft failure or patient death (HR 2.119; P ¼ 0.019), as was the patient cytomegalovirus status (HR 0.459; P ¼ 0.023). None of the Cox regression models employed showed a significant association between FCXM and either BPAR in the first year (HR 0.993; P ¼ 0.987), or graft failure or patient death during the period of follow-up (HR 0.521; P ¼ 0.246). Further Kidney International (2006) 70, 1474–1481

original article


exploratory analyses included modeling the effect of FCXM as a numeric predictor, as well as modeling donor source as a covariate rather than as a stratification variable and including the donor source by FCXM interaction in the Cox model 0.9 Survival function

0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 1000

2000

3000

4000

Survival time (days) Negative FCXM

Positive FCXM

b 1.0 0.9

Survival function

0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 1000

0

2000

3000

4000

Survival time (days) Positive FCXM and one tx

Positive FCXM and > one tx

Negative FCXM and one tx

Negative FCXM and > one tx

Predictive value of FCXM

The predictive value of the T-cell FCXM was examined using both dichotomous and continuous measures of assay response. Employing a value of 10 MCS or more (Table 4), the predetermined diagnostic standard for a positive FCXM, the sensitivity and specificity were 0.128 and 0.883 for the total study population, and the positive and negative likelihood ratios (LRs) were close to 1 at 1.093 and 0.988, respectively. The odds of BPAR were 0.436 (78/179); the positive and negative post-test probabilities (PTPs) were 0.323 and 0.301, respectively. No improvement in the predictive value was observed when patient groups were analyzed separately. For living donor transplant recipients, the sensitivity and specificity were 0.052 and 0.915, respectively, and the positive and negative LRs were close to 1 at 0.612 and 1.036, respectively. The prevalence of BPAR was 0.290 (58/200) for the living donor transplant recipients, and the positive and negative PTPs were therefore 0.200 and 0.297. For the group of patients with cadaver donors, the sensitivity and specificity of the FCXM were 0.350 and 0.757, and the positive and negative LRs were 1.439 and 0.859. The prevalence of acute rejection in this subset of patients was 0.351 (20/57), so that the positive and negative PTPs were 0.438 and 0.317, respectively. Exploration of the quantitative influence FCXM as a linear variable using ROC analysis provided no additional contributory information. The general performance of the FCXM was not significantly different from chance alone, which can be expressed as the diagonal straight line in each ROC plot (P ¼ NS). The area under the curve for living transplant recipients was 0.483,

a 1.0

0

analysis. In all cases, results were very similar to those described above.

c 1.0 0.9 0.7

ROC for cadaveric donor

ROC for living donor

0.6

1.000

0.9

0.900

0.4

0.8

0.800

0.3

0.7

0.700

0.2

0.6

0.600

0.1 0

1000

2000

3000

4000

Survival time (days) Positive FCXM and CAD

Positive FCXM and LD

Negative FCXM and CAD

Negative FCXM and LD

Sensitivity

1

0.5

Sensitivity

Survival function

0.8

0.5 0.4

0.500 0.400

0.3

0.300

0.2

0.200

0.1

0.100 0.000

0

Figure 2 | Kaplan–Meier estimate of time to graft loss or death by (a) FCXM status, (b) by FCXM status and graft number, and (c) by FCXM status and donor source.

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1-Specificity

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1-Specificity

Figure 3 | ROC curve of flow cytometry crossmatch in patients with living donors and patients with cadaver donors. Dotted line depicts line of identity.

Table 4 | Clinical performance of FCXM in renal transplantation: sensitivity and specificity, LRs, and pre/post-test probabilities Patient category Living donor Cadaver donor All patients

Prevalenceee

Sensitivity

Specificity

LR (+)

LR ()

PTL (+)

PTL ()

0.290 0.351 0.304

0.052 0.350 0.128

0.915 0.757 0.883

0.612 1.439 1.093

1.036 0.859 0.988

0.200 0.438 0.323

0.297 0.317 0.301

LR, likelihood ratio; PTL, pre-test probabilities; PTP, post-test probabilities.


1477

original article

(P ¼ NS, Figure 3a), whereas for cadaver transplant recipients it was 0.572 (P ¼ NS, Figure 3b). DISCUSSION

FCXM has been employed as an adjunctive technique to predict graft outcome in renal transplantation for over two decades.11,15,16 During this time, evidence has accumulated to indicate that this assay is more sensitive than conventional CDC techniques, may detect antibodies which are undetected by other routine crossmatch methods, and may provide important predictive information regarding patient risk and outcome.30 However, this information is complicated by clinical, methodological, and analytical differences between reports. Variable inter-laboratory consensus and false-negative rates point to concerns of assay precision and specificity30,31 that may reflect differences in methodology or equipment,30,32–34 in operator technique, or in criteria for defining a positive result in different centers.33,35,36 In the absence of clear evidence founded on prospective-blinded studies, the objective prognostic performance of FCXM remains unclear. There is consequently no universal policy regarding its clinical application or agreement regarding its predictive implications. Use of this technique, and the therapeutic interpretation, remain at the discretion of the transplant program or physician, and clinical response to a positive test range from nil, to the use of intensified immunosuppression, active desensitization, or withholding of the transplant. This study showed that when subjected to rigorous analysis, T-cell FCXM did not provide incremental information adequate to discriminate patients at high risk of acute rejection, graft loss, or death. The patient cohort comprised subjects across a broad range of anticipated risk throughout a 4-year period of care, who were then followed for up to 7 years post-transplant. Although patient selection and management remained reasonably constant during these periods, several important risk factors were unequally distributed between the FCXM- positive and-negative groups. Adjustment for important confounders including donor age, recipient PRA, HLA mismatch, immunosuppression, and the occurrence of DGF after graft implantation37–40 was therefore mandatory in order to determine the incremental predictive value FCXM. Five principal means are employed to control for confounders: restriction, randomization, matching, stratification, and multivariable analysis.41 Stratification and multivariable analysis were selected as the preferred options in this study. Log-rank comparison of Kaplan–Meier estimates showed no difference in the time to BPAR, or graft loss, or death between patients with a positive or negative FCXM. Cox models were stratified for living and cadaver transplant recipients to account for differential risks associated with donor source and immunological status. The models were also fitted separately to the two patient subgroups in an attempt to improve the precision with the achievement of homoscedasticity, and all models were examined with and 1478


without DGF as a covariate. These analyses showed no independent predictive value of the FCXM in the current patient sample. Clinical performance of diagnostic tests is determined by measures of discriminant capacity.42 Concordance with a given disease state is determined by the true-positive (sensitivity) and true-negative (specificity) rates, whereas discordance is reflected by the false -positive and falsenegative rates. Incorporation of disease or event prevalence enables the calculation of post-test probability, which measures the predictive value of the test and indicates the possibility of having the disease after the diagnostic result is known. An ideal diagnostic test would therefore be both sensitive and specific, with a high post-test probability when the result was positive and a low post-test probability when the result was negative.42 The sensitivity and specificity of the T-cell FCXM were 0.13 and 0.88, respectively, indicating that only 13% of patients with biopsy-proven acute rejection had a positive FCXM, whereas 88% of patients without biopsy-proven acute rejection had a negative FCXM. In addition, the positive and negative LRs were close to 1 in both the pooled data and recipients subgroups showing that the capability of FCXM to ‘rule in’ or ‘rule out’ BPAR was very poor. ROC analysis was employed to explore the predictive information obtained at cutoff points higher or lower than the 10 MCS selected as the diagnostic standard of the positive FCXM. This analysis showed no significant relationship with BPAR, and implied that FCXM provided no important gain in information for clinical decision making at other cut-points. The probability of BPAR in the study population (pre-test probability) was 30.4%, consistent with the population incidence of this event during this period of observation. The presence of a positive FCXM increased this risk (posttest probability) only marginally to 32.3%, whereas a negative FCXM decreased the probability to 30.1%, providing little discriminant clinical value. Because of the reported importance of FCXM for re-grafted and presensitized patients,19,20 the performance of FCXM was also assessed separately for each subgroup of patients. FCXM offered no additional predictive value in living DRs, and in fact a positive test was associated with a decreased risk of BPAR. Among high-risk cadaveric graft recipients in whom the pre-test probability of BPAR was 35.1%, a positive FCXM increased the post-test probability of BPAR to 43.8% whereas a negative test decreased the post-test probability marginally to 31.7%. Post-test probability is directly related to the population prevalence of disease or event, and declines rapidly as this prevalence decreases.3,42 We therefore applied the pooled test results of this study to current practice, assuming a pre-test rate of BPAR ranging from 5 to 15% as reported in recent prospective studies.43–47 Positive (FCXM þ ) and negative (FCXM ) PTPs of BPAR across this range would be 5.4 and 4.9%, respectively at the lower extreme, rising to 16.2 and 14.8% at the upper border. Even when the results of the high-risk cadaver group are applied to the same pre-test Kidney International (2006) 70, 1474–1481


range for BPAR, there is little evidence of important discriminant value: positive and negative PTPs ranges from 7.0 to 4.3%, respectively at a pre-test BPAR of five, to 20.3 and 13.2% at a pre-test BPAR of 15%. This narrow range provides little incremental value for clinical decision formulation. To ensure a high degree of risk distribution, the study population was designed to include both low-risk living donor and immunologically high-risk cadaver graft recipients throughout the period of study. Low-risk primary cadaver transplant recipients were therefore under-represented in the analysis by comparison with the total transplant population, and the ability of FCXM to predict acute rejection in such patients is therefore less clear. For this reason, statistical modeling was undertaken incorporating all 493 patients transplanted during the 4 years of study, and allocating a plausible range of outcomes between those exhibited by the live donor and high-risk cadaver DRs. This analysis showed no improvement in predictive value, even when all additional patients were considered comparable to the high-risk cadaver recipient group examined here (P ¼ 0.480). The T-cell FCXM assay was performed using a fluorescence-activated cell sorter with a 256 channel scale. The threshold for a positive test was o10 MCS, comparable to a value of 30–40 MCS using a 1024 channel scale. It is possible that this cut-point is either too high or low, thereby impacting the false-negative or false-positive rates. For this reason we have analyzed the data in two additional ways. First, in the Cox models stratified by donor source, we have treated the FCXM result as a dichotomous variable (i.e. o10 MCS or p10 MCS) and as a continuous variable (i.e. 0-maximum value recorded). Second, we have employed a receiver operating characteristic (ROC) curve analysis for the combined and distinct (LD, DD) data to evaluate the point of inflection, if such existed for any data set. As reported in the text, neither of these approaches suggested either a clinically or statistically significant influence of the FCXM assay at any threshold. In summary, this prospective, blinded study examined he ability of T-cell FCXM to discern those recipients with high risk of rejection, or graft loss or death. Based on its measured performance, the FCXM was unable to provide clinically important information to indicate a change in the selection for transplantation. Kaplan–Meier analysis indicated that the probability of acute rejection was not statistically different in patients with a positive or negative FCXM, and ROC models showed that the magnitude of FCXM had no effect on the result obtained. Multivariate Cox analysis used to control for confounders between study groups and to adjust for important covariates showed no significant correlation between test and clinical outcome. Finally, the PTPs did not provide clinically important guidance in the selection of patients according to subsequent risk. In concert, these analyses suggest that FCXM not an independent predictor of outcome and its application in predicting short-term or longterm risk,48–50 or in determining patient selection or Kidney International (2006) 70, 1474–1481

original article

immunosuppressive treatment may need to be re-evaluated. It is likely that newer solid-phase assays, which demonstrate comparable analytical sensitivity with greater analytical specificity and hence exhibit a lower false-positive rate, may supplant the FCXM in guiding patient selection or therapy, but the clinical utility of these technologies remains to be proven by rigorous prospective studies. MATERIALS AND METHODS Hypothesis The study was conducted to test the hypothesis that T-cell FCXM before transplantation in patients with a negative anti-donor AHGXM provides independent incremental prediction of graft rejection or long-term outcome under conditions of normal clinical practice. Patients This prospective cohort study included 257 patients who underwent renal transplantation between January 1997 and December 2000 at the University of British Columbia, Vancouver, Canada. The study included both living donor and cadaver graft recipients across a spectrum of immunological risk. All donors and recipients were ABO compatible, and all recipients had a negative donor T celldirected AHG-XM. FCXM was performed before transplantation, but the results were not provided to the clinical team and did not affect patient selection or treatment decisions. Patients were followed in the University of British Columbia transplant program up to the point of graft loss or death, and there was no loss to follow-up during the period of observation. Treatment Before 1998 most patients received immunosuppression consisting of cyclosporine, azathioprine, and prednisone. In 1998, immunosuppression was changed to cyclosporine, prednisone, and mycophenolate mofetil for the first transplant year, then maintained as cyclosporine, azathioprine, and prednisone in the following years. During this period of time 60 patients received tacrolimus instead of cyclosporine; 12 patients received sirolimus; and 20 patients at high immunological risk received induction therapy with OKT3. FCXM T-cell FCXM was performed on fresh pre-transplant recipient serum collected for conventional CDC. Donor lymphocytes were obtained by gradient separation from the peripheral blood or donor spleen and re-suspended in media. One million cell aliquots were transferred into polystyrene tubes and pre-incubated with goat Ig at 150 mg/tube (Caltag Laboratories Ltd, Burlington, CA, USA). Negative control, positive control (pooled patient sera 490% PRA), or patient’s sera (100 ml) were added to the cells and incubated at 41C for 30 min. Cells were washed twice before the addition of 3.75 mg fluorescein isothiocyanate-conjugated anti-human IgG-Fcspecific reagent to each tube (Jackson Research Laboratory Inc. West Grove, PA, USA). They were incubated for 5 min followed by the addition of 0.125 mg PerCP-conjugated anti-CD3 per tube (Beckton Dickinson, San Jose, CA, USA). The tubes were further incubated at 41C for 30 min. The cells were then washed, fixed in 1% para formaldehyde and analyzed by flow cytometry using a fluorescenceactivated cell sorter (Becton Dickinson) with a 256 channel log scale. T-lymphocytes were gated according to their forward scatter and side scatter properties and the positivity of the CD3 surface antigens. 1479

original article

The crossmatch result was determined by the difference in the mean fluorescence of the bound Fc-specific anti-IgG in the crossmatch and the mean fluorescence of the bound Fc-specific anti-IgG in the negative control, based on a double gating strategy, and a positive T-cell FCXM was defined by a channel shift of p10 based on statistical analysis of normal control distributions.51


McManus, and Dr Rob McMaster for their contribution to and critical review of the manuscript.

REFERENCES 1. 2.

Definition Acute rejection episodes were diagnosed according to standard clinical and laboratory criteria, and were confirmed by graft biopsy. Acute rejection episodes were treated with intravenous pulse methylprednisolone, and steroid-resistant episodes then received either OKT3 or antilymphocyte globulin. Biopsy-proven or treated acute rejection episodes were analyzed independently. DGF was defined as a delay in the fall of serum creatinine with the need for hemodialysis within the first week following transplantation. Sensitivity meant the proportion of recipients with acute rejection who were correctly identified by a positive FCXM. Specificity meant the proportion of recipients without acute rejection who were correctly identified by a negative FCXM. Data analysis Categorical variables were summarized using frequency and percentages whereas continuous variables were summarized using the median and first and third quartiles, [Q1, Q3], unless otherwise indicated. Times to acute rejection during the first year posttransplant, and graft loss, or death at any time were analyzed using survival methods. Patients not experiencing these events were censored at the latest point of follow-up. The log-rank test was used to compare Kaplan–Meier estimates of BPAR in patients with a positive (10 MCS or greater) and negative FCXM, with and without stratification for transplant number and donor source. The combined effects of multiple risk factors were explored using Cox proportional hazard models stratified by donor type (cadaveric vs living donor). Covariates in the models included the T-cell FCXM, PRA level (PRA), HLA-A, B mismatch number, number of HLA-DR mismatches, number of prior transplants, immunosuppression, donor and recipient ages, and cytomegalovirus serum statuses. DGF was treated as a contingent risk factor, being both a consequence of sensitization and cause of graft rejection/loss, and analysis was performed with and without DGF as a covariate. Exploratory backwards elimination was also used to identify important predictors; however, P-values and HRs were based on the full multivariate regression model. ROC curves were drawn to show the general performance of FCXM. Sensitivity and specificity were calculated at the cut-point of 10 MCS of FCXM. Positive (PTP þ ) and negative PTPs (PTP) were calculated to measure the predictive value of FCXM.3,28 The positive LR ( þ ) and the negative LR () provide clear measurements of the ‘ruling-in’ and ‘ruling out’ capabilities of a test. A LR ( þ ) exceeding 10 was considered reasonable evidence that acute rejection would be highly likely happen when the FCXM was positive; 2–5 was considered poor to fair. A LR () less than 0.1 was considered reasonable evidence to rule out acute rejection, whereas 0.2–0.5 was poor to fair. P was the prevalence of acute rejection. ACKNOWLEDGMENTS

Dr Rongzhu Wen was supported by a fellowship from the International Society of Nephrology. We are indebted to the members of the Genome Canada Biomarkers in Transplantation Group: Dr David Landsberg, Dr R Jean Shapiro, Dr John Gill Dr Bruce 1480

3. 4. 5. 6.

7.

8. 9.

10. 11.

12. 13.

14.

15. 16.

17.

18. 19.

20.

21.

22. 23.

24.

25.

26.

27.

Group BDW. Biomarkers and surrogate endpoints: preferred definitions and conceptual framework. Clin Pharmacol Ther 2001; 69: 89–95. Mayeux R. Biomarkers: potential uses and limitations. Neurorx 2004; 1: 182–188. LaBaer J. So, you want to look for biomarkers. J Proteome Res 2005; 4: 1053–1059. Morris P, Monaco A. Antibody revisited. Transplantation 2004; 78: 179–180. Bohmig G, Regele H. Diagnosis and treatment of antibody-mediated kidney allograft rejection. Transpl Int 2003; 16: 773–787. Knight RJ, Burrows L, Bodian C. The influence of acute rejection on long-term renal allograft survival: a comparison of living and cadaveric donor transplantation. Transplantation 2001; 72: 69–76. Knight RJ, Burrows L. The combined impact of donor age and acute rejection on long-term cadaver renal allograft survival. Surgery 1999; 125: 318–324. Boratynska M, Szepietowski T, Szewczyk Z. Acute rejection and delayed graft function–risk factors of graft loss. Ann Transplant 1996; 1: 19–22. Hariharan S, Alexander JW, Schroeder TJ et al. Impact of first acute rejection episode and severity of rejection on cadaveric renal allograft survival. Clin Transplant 1996; 10: 538–541. Gebel HM, Bray RA. Sensitization and sensitivity: defining the unsensitized patient. Transplantation 2000; 69: 1370–1374. Garovoy MG, Rheinschmidt MA, Bigos M et al. Flow cytometry analysis: a high technology crossmatch technique facilitating transplantation. Transplant Proc 1983; 15: 1939. Leenaerts PL, De Ruysscher D, Vandeputte M et al. Measurement of alloantibody by flow cytometry. J Immunol Methods 1990; 130: 73–79. Scornik JC, Brunson ME, Schaub B et al. The crossmatch in renal transplantation. Evaluation of flow cytometry as a replacement for standard cytotoxicity. Transplantation 1994; 27: 621–625. Cook DJ, Fettouh HI, Gjeetson DW et al. Flow cytometry crossmatching in the UNOS Kidney Transplant Registry. In: Terasaki P, Cecka JM (eds). Clinical Transplants. UCLA Immunogenetics Center: Los Angeles, 1998, pp 413–419. Harmer AW. Utilization of crossmatch techniques for renal transplantation. Curr Opin Nephrol Hypertens 1998; 7: 687–690. Gebel H, Bray R, Nickerson P. Pre-transplant assessment of donor-reactive, HLA-specific antibodies in renal transplantation: contraindication vs. risk. Am J Transplant 2003; 3: 1488–1500. Takemoto S, Zeevi A, Feng S et al. National conference to assess antibody-mediated rejection in solid organ transplantation. Am J Transplant 2004; 4: 1033–1041. Fuggle S, Martin S. Toward performing transplantation in highly sensitized patients. Transplantation 2004; 78: 186–189. O’Rourke R, Osorio R, Freise C et al. Flow cytometry crossmatching as a predictor of acute rejection in sensitized recipients of cadaveric renal transplants. Clin Transplant 2000; 14: 167–173. Kerman RH, Van Buren CT, Lewis RM et al. Improved graft survival for flow cytometry and antihuman globulin crossmatch-negative retransplant recipients. Transplantation 1990; 49: 52–56. Karpinski M, Rush D, Jeffery J et al. Flow cytometric crossmatching in primary renal transplant recipients with a negative anti-human globulin enhanced cytotoxicity crossmatch. J Am Soc Nephrol 2001; 12: 2807–2814. Utzig MJ, Blumke M, Wolff-Vorbeck G et al. Flow cytometry cross-match: a method for predicting graft rejection. Transplant Proc 1997; 63: 551. Ogura K, Terasaki P, Johnson C et al. The significance of a positive flow cytometry crossmatch test in primary kidney transplantation. Transplantation 1993; 56: 294–298. Pelletier RP, Orosz CG, Adams PW et al. Clinical and economic impact of flow cytometry crossmatching in primary cadaveric kidney and simultanoeus pancreas-kidney transplant recipients. Transplantation 1997; 63: 1639–1645. Kerman RH, Susskind B, Buyes I et al. Flow cytometry-detected IgG is not a contraindication to renal transplantation. Transplant Proc 1999; 68: 1855. Evans PR, Lane AC, Lambert CM et al. Lack of correlation between IgG T-lymphocyte flow cytometric crossmatches with primary renal allograft outcome. Transplant Int 1992; 5: S609–S612. Christiaans MHL, Overhof R, ten Haaft A et al. No advantage of flow cytometry crossmatch over complement-dependent cytotoxicity in Kidney International (2006) 70, 1474–1481

original article


28. 29.

30.

31.

32.

33.

34.

35.

36.

37.

38.

39.

immunologically well-documented renal allograft recipients. Transplantation 1996; 62: 1341–1347. Schechter MT, Sheps SB. Diagnostic testing revisited: pathways through uncertainty. Can Med Assoc J 1985; 132: 755–760. Pincus M. Interpreting laboratory results: reference values and decision making. In: Henry J (ed). Clinical Diagosis and Management by Laboratory Methods, 19th edn. WB Saunders Company: Philadelphia, 1996, pp 76–77. ASHI. ASHI Proficiency Testing Program. Summary Report, Survey 2004 AC-2. HLA antibody screening/crossmatching, in, American Society of Histocompatibility and Immunogenetics, 2004 pp 1–79. CAP. Surveys 2003: Class I Crossmatching, PRA and Antibody identification, in, College of American Pathologists: Northfield, Il, USA, 2003 pp 1–16. Lobo PI, Isaacs RB, Spencer CE et al. Improved specificity and sensitivity when using pronase-digested lymphocytes to perform flow-cytometric crossmatch prior to renal transplantation. Transplant Int 2002; 15: 563–569. Cho YW, Cecka JM. Crossmatch tests-an analysis of UNOS data from 1991–2000. In: Cecka JM, Terasaki P (eds). Clinical Transplants, 2001: 237–246. Vaidya S, Cooper TY, Avandsalehi J et al. Improved flow cytometric detection of HLA alloantibodies using pronase: potential implications in renal transplantation. Transplantation 2001; 71: 422–428. Harmer AW, Garner S, Bell AE et al. Evaluation of the flow cytometric crossmatch: preliminary results of a multicenter study. Transplantation 1996; 61: 1108–1111. Shenton BK, Bell AE, Harmer AW et al. Importance of methodology in the flow cytometric crossmatch: a multicentre study. Transplant Proc 1997; 29: 1454–1455. Mota A, Figueiredo A, Cunha MF et al. Risk factors for acute rejection in 806 cyclosporine-treated renal transplants: a multivariate analysis. Transplant Proc 2003; 35: 1061–1063. McLaughlin K, Wu C, Fick G et al. Cytomegalovirus seromismatching increases the risk of acute renal allograft rejection. Transplantation 2002; 74: 813–816. Kandaswamy R, Humar A, Payne WD et al. Risk factors for kidney transplant acute rejection: a multivariate analysis. Transplant Proc 2001; 33: 1112.


40.

41. 42. 43.

44.

45.

46.

47.

48.

49.

50.

51.

Meier-Kriesche HU, Steffen BJ, Hochberg AM et al. Long-term use of mycophenolate mofetil is associated with a reduction in the incidence and risk of late rejection. Am J Transplant 2003; 3: 68–73. Daya S. Characteristics of good causation studies. Semin Rep Med 2003; 21: 73–83. Sox HJ, Blatt M, Higgins M et al. Medical Decision Making. Butterworth-Heinemann: Boston, 1988. Flechner S, Goldfarb D, Modlin C et al. Kidney transplantation without calcineurin inhibitor drugs: a prospective, randomized trial of sirolimus versus cyclosporine. Transplantation 2002; 74: 1070–1076. Flechner S, Feng J, Mastroianni B et al. The effect of 2-gram versus 1-gram concentration controlled mycophenolate mofetil on renal transplant outcomes using sirolimus-based calcineurin inhibitor drug-free immunosuppression. Transplantation 2005; 79: 926–934. Gonwa T, Mendez R, Yang H et al. Randomized trial of tacrolimus in combination with sirolimus or mycophenolate mofetil in kidney transplantation: results at 6 months. Transplantation 2003; 75: 1213–1220. Hardinger K, Bohl D, Schnitzler M et al. A randomized, prospective, pharmacoeconomic trial of tacrolimus versus cyclosporine in combination with thymoglobulin in renal transplant recipients. Transplantation 2005; 80: 41–46. Vitko S, Klinger M, Salmela K et al. Two corticosteroid-free regimens-tacrolimus monotherapy after basiliximab administration and tacrolimus/mycophenolate mofetil-in comparison with a standard triple regimen in renal transplantation: results of the atlas study. Transplantation 2005; 80: 1734–1741. Talbot D, Cavanagh G, Coates E et al. Improved graft outcome and reduced complications due to flow cytometric crossmatching and DR matching in renal transplantation. Transplantation 1992; 53: 925–928. Bryan CF, Baier KA, Nelson PW et al. Long-term graft survival is improved in cadaveric renal retransplantation by flow cytometric crossmatching. Transplantation 1998; 66: 1827–1832. El Fettouh HA, Cook DJ, Bishay E et al. Association between a positive flow cytometry crossmatch and the development of chronic rejection in primary renal transplantation. Urology 2000; 56: 369–372. Stephenson M, Wu V, Mackinnon M et al. Standardization of flow cytometric crossmatch (FCXM) for investigation of unexplained habitual abortion. Am J Reprod Immunol 1995; 33: 1–9.

1481