Treatment of relapsed leukemia after unrelated donor marrow transplantation with unrelated donor leukocyte infusions

Patients were identified from the NMDP database. All centers that requested donor leukocyte products through the NMDP for “relapse” or “secondary malignancy” were contacted in writing; 19 of 20 centers contacted agreed to participate, accounting for 58 of 61 patients (95%) who received UDLI for relapse organized through the NMDP. Questionnaires were sent to each center designed to assess pretransplant characteristics and complications, relapse characteristics, and details of UDLI therapy. Most data were collected from individual centers, but some data, such as demographics and donor information, were provided through NMDP records. Data collection included, but was not limited to, the following: (1) Patient demographics: age and sex; (2) Pre-BMT characteristics: diagnosis, cytogenetic abnormalities, pre-BMT therapy; (3) BMT characteristics: time from diagnosis to BMT, disease status at BMT, conditioning regimens, complications of BMT (including acute and chronic graft-versus-host disease and other complications); (4) Disease characteristics: disease status at relapse, time from BMT to relapse, treatment for relapse other than UDLI; (5) UDLI characteristics: indication for UDLI, time from relapse to UDLI, performance status, chimerism analysis, use of chemotherapy before UDLI, cell dose administered and number of infusions given, concurrent use of cytokines, modification of donor leukocyte product; (6) Response to UDLI: overall response, use of additional therapy (ie, cytokines), toxicity including acute and chronic GVHD, pancytopenia and marrow aplasia, infections, chimerism after UDLI, treatment and response of GVHD, current disease status, date and cause of death.

UDLI was performed from October 1993 through February 1997; 58 patients received UDLI for relapsed disease, and 30 patients received UDLI after a T-cell– depleted transplant. Patient characteristics are summarized in Table 1.

For patients with CML, a complete remission (CR) was defined as a complete cytogenetic response. Responses were scored as partial response (more than 50% reduction in abnormal cytogenetics), complete cytogenetic response (no detectable cells containing the Philadelphia chromosome on at least 1 occasion), and complete molecular response (negative polymerase chain reaction [PCR] assay for bcr/ablmessenger RNA [mRNA]). For this analysis, 1 negative PCR test forbcr/abl was considered a complete molecular response. Of 8 patients who were classified as achieving a molecular remission, more than 1 negative PCR test was documented in 5 patients. Two patients received UDLI while in a cytogenetic remission but with evidence of disease by PCR (molecular relapse); CR in these patients was defined as a “molecular remission.” Early-phase relapse of CML included patients with molecular, cytogenetic, or chronic phase relapse. Patients were considered to have late-phase relapse of CML if they were treated for accelerated or blast-phase CML. Patients with accelerated-phase CML based on abnormal cytogenetics were initially classified as having advanced-phase CML.

For patients with diseases other than CML, responses were scored as CR (less than 5% blasts in bone marrow or no other evidence of disease) or partial remission (more than 50% reduction in bone marrow or peripheral blasts or more than 50% reduction in adenopathy). Patients who received UDLI in conjunction with cytoreductive chemotherapy (n = 4) or after achieving remission from other therapy (n = 7) were considered not evaluable for a direct response to UDLI. The analysis was performed both including these patients (based on disease status after UDLI regardless of pre-DLI therapy) and excluding these patients, as noted below.

Based on data from other DLI series,1 3 a minimum follow-up of 28 days after UDLI was required for patients to be considered evaluable for response, acute GVHD, or marrow aplasia. One patient with grade IV acute GVHD 5 days after UDLI died on day 27 without a response and was considered evaluable for acute GVHD but not response or aplasia. No other patient who survived 28 days or fewer after UDLI developed acute GVHD or marrow aplasia. A minimum follow-up of 100 days after UDLI was required for patients to be evaluable for chronic GVHD.

HLA antigen matching was defined as matched versus mismatched according to NMDP criteria. Mismatched patients include “minor” mismatches (mismatches within cross-reactive antigen groups) and major mismatches (non–cross-reactive antigen mismatches).

Data were analyzed either with the SAS statistical software (SAS Institute, Cary, NC), or with Statview statistical software package (Abacus Concepts, Berkeley, CA). Overall survival was calculated from the time of first UDLI until death from any cause or last follow-up. Disease-free survival (DFS) was calculated from the time CR was documented until relapse, death, or last follow-up. Factors that were examined for an association with these outcomes included age, sex, donor-patient sex match, HLA-match grade, time from BMT to relapse, time from relapse to UDLI, time from transplant to UDLI, acute and chronic GVHD after transplant, T-cell–depleted graft for original BMT, and indications for UDLI. The probabilities of overall survival and DFS were calculated according to the method of Kaplan and Meier.18 The effects of categorical variables on survival and DFS were examined using the log-rank test,19 and the effects of continuous variables were analyzed with a Cox proportional hazards model.20 Optimal Cox models for survival and DFS were developed using stepwise regression.21 Only factors that were significant at the 0.05 level in the univariate analysis of individual effects were considered in a multivariable analysis. Factors associated with survival and DFS in the multivariable analysis were corrected to adjust for a diagnosis of acute myelogenous leukemia (AML) using Cox regression analysis. Other outcomes examined with a logistic regression model22 included response and incidence and severity of acute GVHD. For patient subsets with smaller sample sizes, the Fisher exact test was used to evaluate associations.23 

In most cases, donor leukocytes were collected by leukapheresis and procured through requests to the NMDP. In several cases, donor T-cells were cryopreserved after T-cell depletion of the donor marrow at the time of transplantation. Cell dose measurements were reported differently by different centers. Cell dose was not reported for 4 patients. The total number of mononuclear cells (MNC) administered was determined for 50 patients, and the CD3⁺ cell content of the donor product was determined in 33 cases, as shown in Table 1. In 28 patients (48%), the cell dose was more than 1 × 10⁸ MNC/kg, and in 24 patients (41%), the cell dose was less than 1 × 10⁸ MNC/kg. The donor product was depleted of CD8⁺ cells in 5 patients, red cells depleted in 4 patients, and gene insertion (herpes simplex virus thymidine kinase) attempted in 1 donor product.

Donor cells were given in 1 infusion in 39 patients (67%), in 2 infusions in 12 patients (21%), and in 3 to 6 infusions in 7 patients (12%). Of the 19 patients who received more than 1 UDLI, the dates of subsequent infusions were available for 17 cases. UDLI was completed in 8 cases within 1 month and, in 15 cases, within 4 months. Although subsequent infusions contained more than 2-fold higher cells than the first infusion (dose escalation) in 8 patients, only 4 patients received at least a 2-fold dose escalation over a period of more than 1 month, and none of these patients achieved a CR. The other patients who received cells over more than 1 month received similar or even lower cell doses.

Of the 58 patients treated with UDLI for relapsed disease, 8 were in remission prior to UDLI. Of the 50 patients included in the response analysis, 21 (42%; 95% confidence interval [CI], 28%-56%) were in CR after UDLI, 4 patients (8%) achieved a partial remission (PR), and 22 (44%) had no response (Table 2). Follow-up was insufficient to assess response in 3 patients.

Nineteen patients (33%) are alive (17 in CR), with a median follow-up time from UDLI of 66 weeks (range, more than 9-180 weeks). Fifteen patients (26%) remain in remission more than 9 to 155 weeks (median, 66 weeks) after documented CR, 2 are in CR but have minimal residual disease detected only by PCR, and 2 are alive with active disease. The probability of DFS for patients with CML, AML, and acute lymphoblastic leukemia (ALL) who were in CR after UDLI is shown in Figure 1. Seven patients who achieved a CR relapsed with their original disease and subsequently died.

In prior analyses of family-member DLI, the treatment was most effective for patients with CML. Similarly, in this series, 11 of 24 (46%; 95% CI, 26%-66%) recipients of UDLI for relapsed CML were in remission after UDLI (1 additional patient was in CR prior to UDLI). Eleven patients had no response, and 2 patients had a PR (Table 2). The probability of DFS for the 12 patients in CR after UDLI is shown in Figure 1; 8 of the 12 patients remain alive in CR more than 17 to 155 weeks (median, 67 weeks) after documented remission. The median survival after UDLI for patients treated for relapsed CML was 42 weeks. Eleven patients (44%) remain alive more than 10-180 weeks (median, 79 weeks) after UDLI. This includes 7 patients treated for early-phase CML and 4 patients treated for advanced-phase CML.

Of 12 patients treated for “early-phase” relapse, 7 (58%) achieved a CR; CR was documented by PCR in 5 cases and by cytogenetics without PCR in 2 cases. The treatment characteristics and outcome for these 12 patients according to disease status are summarized in Table3.

Thirteen patients were treated for advanced-phase disease (accelerated phase [n = 8] or blast crisis [n = 5]), including 2 patients who relapsed with advanced disease but received chemotherapy before UDLI. The median cell dose administered to these 13 patients was 0.9 × 10⁸ MNC/kg (range, 0.17 × 10⁸ to 5 × 10⁸ MNC/kg). Five of the 13 patients were in CR after UDLI; CR was documented by PCR in 3 patients and cytogenetics in 2 patients. Four of these patients had a direct response to UDLI, and 1 was in CR prior to UDLI. All 4 patients with a direct response to UDLI were treated for accelerated phase and remain alive more than 48 to 155 weeks after documentation of CR without a relapse. No patient with blast-crisis CML had a direct response to UDLI with a minimum follow-up of 5 weeks, and all 5 patients died 5 to 17 weeks after UDLI from progressive disease (n = 4) or infection (n = 1).

Because many patients who relapse after allogeneic BMT will have complex karyotype abnormalities but have a clinical course typical for chronic-phase CML, we analyzed response rates classifying patients with accelerated-phase CML based only on cytogenetic abnormalities as “early-phase relapse” and classified only patients with other clinical findings of accelerated phase as “advanced-phase relapse.” In this analysis, 10 of the 17 patients (59%) with early-phase relapse were in CR after UDLI, and 2 of the 8 patients with advanced-phase CML were in CR after UDLI. The overall survival appears superior for recipients of UDLI for early-phase relapse of CML compared with patients with advanced-phase relapse (P = .02) (Figure2) when these criteria are used for the analysis.

All 5 patients treated for relapsed CML (early-phase relapse [n = 4]; advanced-phase relapse [n = 1]) who received donor products depleted of CD8⁺ cells achieved a CR and remain in remission 17 to 135 weeks after UDLI.

AML is less responsive to family-member DLI. Four patients with AML were in CR before UDLI. Of 19 patients not in CR prior to UDLI, 8 (42%; 95% CI, 20%-64%) were in CR after UDLI (Table 2). Nine patients had no response, 1 patient had a PR, and follow-up was insufficient in 1 patient. Four of the 12 patients in remission after UDLI eventually relapsed 8 to 31 weeks after documentation of remission; all 4 patients died from progressive disease. An additional 4 patients in CR after UDLI died from treatment-related complications (GVHD [n = 1]; infection [n = 3]). Five patients (22%) are alive in CR (n = 4) or with persistent cytogenetic evidence of disease (n = 1) more than 9 to 102 weeks (median, 55 weeks) after UDLI. The probability of DFS for patients who were in CR after UDLI is shown in Figure 1, and the median survival after UDLI for relapsed AML patients was 11 weeks.

DLI is typically the least effective in ALL. In this series, 3 patients were in CR before UDLI: 1 relapsed and died of progressive disease, 1 died of GVHD, and 1 remains alive in CR 16 weeks after UDLI. Of the remaining 4 patients, 2 (50%; 95% CI, 1%-99%) were in CR after UDLI (Table 2); 1 remains alive 158 weeks after UDLI, and 1 relapsed 14 weeks after UDLI and died of progressive disease. The median survival after UDLI for relapsed ALL patients was 35 weeks. One patient had no response and died of progressive disease, and 1 patient died within a week of UDLI with insufficient follow-up. The probability of DFS for the 5 patients who were in CR after UDLI is shown in Figure 1.

Three patients with non-Hodgkin's lymphoma (n = 2) or chronic myelomonocytic leukemia (n = 1) failed to achieve CR after UDLI (1 CR, 1 PR, and 1 with follow-up less than 28 days), with a follow-up of 2 to 24 weeks after UDLI. Two patients died from progressive disease, while 1 remains alive with active disease 24 weeks after UDLI.

Fifty-six patients were evaluable for GVHD with follow-up at least 28 days after UDLI, and these data are summarized in Table 2. Sixty-one percent of patients had no GVHD, and grade II-IV acute GVHD developed in 25% of patients. The incidence or severity of acute GVHD was not related to the indication for UDLI, cell dose administered, or a prior history of GVHD after BMT.

Of 32 patients surviving more than 100 days after UDLI, 13 (41%) developed chronic GVHD. Limited chronic GVHD occurred in 3 patients, and extensive chronic GVHD was seen in 10 patients. Seventeen (53%) had no chronic GVHD with a median follow-up of 43 weeks (range, 16-180 weeks) after UDLI, and 2 patients were not evaluable for chronic GVHD.

The incidence of chronic GVHD was not associated with the indication for UDLI, cell dose administered, or a history of GVHD after BMT. In addition, acute GVHD after UDLI did not predict development of chronic GVHD in patients surviving longer than 100 days from UDLI; 5 of 11 patients with acute GVHD developed chronic GVHD, whereas 8 of 21 patients without acute GVHD subsequently had clinical signs of chronic GVHD (P = .7).

Four of 34 (12%) evaluable patients developed marrow aplasia attributable to UDLI after receiving 0.1 × 10⁸ to 2.2 × 10⁸ MNC/kg. Aplasia was associated with a CR in 1 patient with CML, 1 patient had a PR, and no response was noted in 2 patients. Two of these patients died from progressive disease, and 1 died from infection as a result of therapy.

Many patients with AML and ALL received UDLI at the time of a chemotherapy-induced nadir or after achieving remission with chemotherapy. To determine the effect of therapy prior to UDLI, patients with adequate follow-up who were evaluable for a direct response were analyzed separately from patients who received other therapy prior to UDLI.

Fifteen patients were evaluable for a direct response; a CR was documented in 5 patients (33%) a median of 3 weeks (range, 1-57 weeks) after UDLI. Two of these patients relapsed with AML 8 and 31 weeks after a documented CR and died of progressive disease. Two additional patients died of treatment-related complications (GVHD and infection), and only 1 remains alive in remission more than 54 weeks after achieving a CR.

Of the 7 patients in CR after UDLI who either received UDLI at the time of a chemotherapy-induced nadir (n = 3) or after achieving a CR from other therapy (n = 4), 2 ultimately relapsed 23 and 24 weeks after UDLI and died of progressive disease. Two of the remaining 5 patients died of infectious complications, and 3 remain alive in CR more than 9 to 66 weeks after UDLI; the time from UDLI to last follow-up in 1 patient was over twice the time from BMT to relapse (66 weeks vs 25 weeks).

There was no significant difference in survival probability between the patients evaluable for a direct response to UDLI and the patients unevaluable for a direct response (P = .7).

Two of 3 patients evaluable for a direct response achieved a CR 3 weeks after UDLI. One patient relapsed with ALL 11 weeks after achieving CR and died of progressive disease. The second patient remains alive in CR more than 158 weeks after UDLI. The patient with no response died of progressive disease.

Three patients received UDLI after chemotherapy-induced CR: 1 remains alive in CR more than 16 weeks after UDLI, 1 patient died of grade IV acute GVHD, and the third patient relapsed and died of progressive disease within 26 weeks of UDLI. One patient died from disease-related complications 1 week after UDLI and was not evaluable for a response.

The use of pre-UDLI therapy does not appear to produce a noticeable survival advantage when compared to patients who did not receive pre-UDLI chemotherapy.

The MNC dose was reported for 50 of 58 patients. These 50 patients received a median dose of 1 × 10⁸ MNC/kg (range, 0.001 × 10⁸ to 31.8 × 10⁸MNC/kg); CD3 counts were determined for 33 of these patients (median, CD3⁺ cell dose, 0.4 × 10⁸/kg; range, 0.01 × 10⁸ to 5.5 × 10⁸/kg). There was no association between cell dose and achievement of CR, presence or severity of acute GVHD (Figures3 and 4), or chronic GVHD. Similarly, there was no association of MNC dose with survival or DFS.

Thirty-nine of the 58 patients (67%) treated with UDLI had died at the time of this analysis (Table 4). Thirteen patients (22%) died from direct treatment-related causes, including 5 patients (9%) who died from GVHD, 6 patients (10%) who died from infectious causes, and 2 patients (3%) who died from other treatment-related causes. Twenty-six of 58 patients (45%) died from progressive disease or other disease-related complications.

We determined whether various pretreatment characteristics were predictive for CR, acute GVHD, survival, and DFS. No factor could be identified that was associated with response. In addition, as summarized in Tables 5 and 6, no factor was associated with GVHD after UDLI. However, in a univariate analysis, several factors were identified that were associated with an improved probability of survival and DFS, including a longer time interval between BMT and relapse and between BMT and UDLI, the use of a T-cell–depleted marrow graft at original BMT, and a diagnosis of CML. When these factors were examined in a multivariable analysis, improved survival was associated with an interval from BMT to relapse of more than 1 year (P = .0015, hazard ratio 0.28, 95% CI, 0.13-0.61), and improved DFS was associated with an interval from BMT to UDLI of more than 1 year (P = .002, hazard ratio 0.09, 95% CI, 0.02-0.40). More patients with AML relapsed and were treated with UDLI within 1 year of BMT (21 of 23 AML patients) than patients with CML (6 of 25 patients) or ALL (4 of 7 patients). When this analysis was adjusted to account for the larger number of patients with AML receiving UDLI within 1 year of BMT, a time interval from BMT to relapse of more than 1 year remained the most significant predictor of survival (P = .0033), and time from BMT to UDLI of more than 1 year remained the most significant predictor of DFS (P = .003). These differences still may be due to more disease-related deaths in patients treated within 1 year of BMT. In these patients, there was no noticeable difference in treatment-related mortality associated with UDLI given within 1 year of BMT.

When patients with CML were analyzed separately, no factor could be identified that was predictive of response or acute GVHD. In a univariate analysis, an interval from BMT to relapse of more than 1 year was associated with improved survival (P = .024) but not DFS (P = .24). The time from BMT to UDLI was also associated with an improved probability of survival (P = .01). In a multivariable analysis, the time from BMT to UDLI remained predictive for improved survival (P = .009, hazard ratio 0.98, 95% CI, 0.97-0.996) but was not associated with DFS (P = .08). No other factor was identified that predicted survival for CML patients other than CML phase at the time of UDLI, as described above.

For patients who relapse after matched sibling allogeneic BMT, DLI has been used to induce a direct GVT reaction. However, this treatment may be associated with potential morbidity and mortality.14Applying this therapy after unrelated donor BMT might be expected to have even greater risks of treatment-related toxicity, because of the greater major and minor histocompatibility differences between unrelated recipients and donors. Other series of DLI describe results of UDLI in only small numbers of patients,3,8 17 and the efficacy and toxicity of UDLI is largely unknown. Therefore, we have compiled results from 58 of the 61 recipients of UDLI identified through the NMDP database to determine the efficacy and toxicity of this therapy and to determine factors that influence the success of UDLI.

Our analysis demonstrates that UDLI can successfully induce a direct GVT effect for patients with relapsed leukemia. Approximately half of the patients were in remission after UDLI, and most of these patients had a direct response to UDLI without additional chemotherapy. Unfortunately, outcome is limited by toxicity and relapse; only one third of all patients remain alive, and 29% remain in remission; however, 48% of patients who were in CR after UDLI remain alive in CR more than 9 to 180 weeks after UDLI.

Patients with CML have the highest likelihood of response after matched sibling DLI. Interestingly, when unrelated donors are used, the response rate for relapsed CML is similar to the rate expected with family-member DLI. Unfortunately, in neither situation is the treatment for blast-crisis relapse effective. For the purpose of this analysis, a CR for patients with CML was defined as a complete cytogenetic remission. Reversion to normal cytogenetics was considered an appropriate definition of CR because the significance of minimal residual disease detected by PCR alone is of unclear significance in CML. However, the remission was further documented in 8 of 12 patients by PCR analysis for thebcr/abl translocation. Although it is not known if all patients achieved a molecular remission, it is notable that in other series of DLI, 95% to 100% of patients who were in cytogenetic remission had no detectable bcr/abl RNA transcripts when studied by PCR.1,3,8 It should be noted that patients were considered to be in molecular CR if they had at least 1 negative PCR test forbcr/abl, although the clinical significance of 1 negative test is not well defined. After allogeneic BMT, sequential positive PCR tests predict for clinical relapse, although some patients have residual disease by PCR detection and do not relapse.24-27Clearly, continued follow-up will be important to determine the significance of positive PCR testing for bcr/abl after UDLI.

Response rates in patients with acute leukemia to matched sibling DLI have generally been poor; only 10% to 20% of patients treated for relapsed AML or ALL achieve CR. In contrast, 42% of patients with AML were in CR after UDLI, and 2 of 4 evaluable patients treated for ALL who were evaluable for a response achieved a CR. The small numbers of patients treated in each group preclude comparisons among disease categories or to matched sibling DLI and are insufficient to allow detailed analysis of factors predictive for response and outcome; however, these response rates are favorable compared with anticipated outcomes for alternative therapy such as second unrelated donor BMT.15,28 29 Although UDLI may be an attractive alternative to second unrelated donor BMT, long-term outcome in either case is unsatisfactory, demonstrating the need to develop newer, more effective approaches to relapse of acute leukemia.

Morbidity and mortality remain significant after UDLI. However, the incidence of acute or chronic GVHD and marrow aplasia may be acceptable compared with other potential treatment options such as second marrow transplantation. The treatment-related mortality rate was 22% in our series, mostly related to GVHD and infections. Although it is difficult to compare results from this retrospective analysis to results from HLA-matched sibling DLI, the incidence and severity of acute and chronic GVHD did not appear to be higher after UDLI compared with large retrospective analyses of DLI from matched siblings.3,8 The incidence of acute GVHD after UDLI was 38% compared with 60% after sibling DLI.3,8 Similarly, the incidence of chronic GVHD was 41% compared with 61% reported after matched sibling DLI.3 However, patients who relapse after unrelated donor BMT may be a selected group with a lower risk of GVHD. The most susceptible patients to severe GVHD may have died or developed severe GVHD after BMT and may have been considered inappropriate candidates for UDLI. Moreover, the graft-versus-leukemia (GVL) effects after unrelated donor BMT may be greater than after matched sibling BMT, even when the marrow graft is T-cell depleted.30 Thus, patients who do relapse may represent a selective population with the lowest risk of GVHD.

Although toxicity after UDLI seems acceptable, strategies to minimize treatment-related morbidity and mortality will be useful. For instance, 10% of patients died of infectious complications directly related to therapy, including marrow aplasia. After matched sibling DLI, the administration of additional donor marrow has successfully reversed marrow aplasia1,4; the early administration of additional donor stem cells, or the use of granulocyte colony-stimulating factor–mobilized donor MNC products, may serve to limit complications related to neutropenia. Several patients received manipulated donor products in an effort to reduce GVHD. All 5 recipients of CD8⁺ cell–depleted products for relapsed CML achieved remissions, similar to results noted after CD8⁺cell–depleted matched sibling DLI.31 32 One of these patients developed grade II acute GVHD, and 2 patients developed chronic GVHD; given the small numbers of patients, the value of this manipulation is unknown.

We were unable to identify any factors that were associated with either response or GVHD after UDLI. Interestingly, in contrast to studies by Mackinnon et al in family-member DLI,9 there was no association of cell dose with response, survival, or toxicity. No patient had a CR after receiving cell doses below 0.1 × 10⁸ MNC/kg, while many patients failed to respond to doses above 1 × 10⁸ MNC/kg. Because there was no association between cell dose and GVHD, it is reasonable to administer 0.1 × 10⁸ to 1 × 10⁸ MNC/kg unrelated donor leukocytes for relapsed disease. In 33 cases, CD3⁺ cell doses were known, and there was a similar lack of association with CD3⁺ cell dose and response, toxicity, and survival. A threshold dose of effector cells may be required for GVHD and GVL activity, and doses above this threshold may not result in quantitative differences in GVHD or GVL activity. Given the lack of association between MNC dose and toxicity, it is also unclear that lower cell doses will result in safer therapy.

Although 19 patients received more than 1 donor cell infusion, a “dose escalation” strategy was unlikely to be an important factor in this analysis. Subsequent infusions were completed within 1 month in 8 patients, and dates of subsequent infusions were not available for 2 patients. Eight of the remaining 9 patients received infusions within 5 months, and 1 patient received an additional UDLI 1 year later. Only 3 patients who received multiple doses of UDLI more than a month apart eventually achieved a CR; in these 3 patients it is difficult to know the cell dose responsible for remission induction. For instance, a response may have occurred from the first infusion if sufficient time was given before subsequent infusions. It is notable, however, that in these 3 patients subsequent doses were similar to the original dose. Given the limited number of patients who received dose escalations over more than 1 month, there is insufficient data to determine a dose-response relationship by this strategy, as has been observed after escalating doses of matched sibling DLI.9 

Intervals from BMT to relapse and BMT to UDLI were the only factors predictive of improved survival and DFS in both a univariate and multivariable analysis of UDLI. There was no significant difference in treatment-related death in the recipients of UDLI within 1 year of BMT, implying that the difference in survival and DFS is not due to excessive toxicity of UDLI when given within a year of BMT. The differences were largely accounted for by an increased incidence in disease-related death in the group of patients treated within 1 year of BMT. Notably, many more patients with AML relapsed and received UDLI within 1 year of BMT than patients with CML or other diseases, which could partially account for the worse prognosis in this group of patients. It is also probable that patients who relapse within 1 year of BMT have more aggressive disease and are least likely to respond to GVL induction. However, given the excessive toxicity anticipated from second BMT within 1 year of the initial transplant, a trial of UDLI or other investigational therapy seems warranted.

In the matched sibling setting, low doses of DLI (for instance, used in the treatment of posttransplant EBV-related lymphoproliferative disorders are felt to minimize the risk of GVHD and aplasia. However, even lower doses of T cells may precipitate GVHD when the histocompatibility disparity between donor and recipient is higher, such as in the unrelated donor transplant setting. In this series, clinically important acute GVHD occurred with low doses of donor lymphocytes, and there was no correlation of cell dose with toxicity. It is therefore possible that even lower doses of MNC will be necessary to further minimize GVHD. It should be noted, however, that factors other than cell dose may have a significant impact on development of GVHD. The degree of minor histocompatibility antigen disparity is likely to influence GVHD. In addition, the timing of DLI may be critical, and infusions given shortly after BMT may result in significantly more GVHD than infusions given with a greater delay after BMT.33 34 

There are several limitations to this study. It is a retrospective analysis of data collected from 19 transplant centers. Treatment strategies differed among institutions, and methods and timing of follow-up testing were not standardized. Both the timing of cytogenetic or PCR testing and the sensitivity of the assays may have varied among laboratories. Reporting bias is also a common concern in this type of analysis, but it is important to note that data were available for 95% of all patients who received UDLI identified through the NMDP database, and it is therefore likely that this study includes a representative population of UDLI recipients. Despite these limitations, it is unlikely that a prospective study will be done soon to address these issues. This study gives a reliable estimate of the value of UDLI and highlights its limitations and areas for future investigation.

The following centers contributed to this study by submitting data: MD Anderson Cancer Center, Houston, TX; Memorial Sloan Kettering Cancer Center, New York, NY; Fred Hutchinson Cancer Research Center, Seattle, WA; Baylor University Medical Center, Dallas, TX; Medical College of Wisconsin, Milwaukee, WI; City of Hope Medical Center, Duarte, CA; Children's Hospital of Philadelphia, Philadelphia, PA; Fairview University Medical Center, Minneapolis, MN; Brigham and Women's Hospital, Boston, MA; Emory University, Atlanta, GA; St. Jude Children's Research Center, Memphis, TN; Methodist Hospital of Indiana, Indianapolis, IN; Children's National Medical Center, Washington, DC; University of Iowa, Iowa City, IA; University of Arkansas, Little Rock, AK; Children's Hospital of Orange County, Orange, CA; Medical College of Virginia, Richmond, VA; Mount Sinai Medical Center, New York, NY; and University of Rochester, Rochester, NY. RHC acknowledges support from the Leukemia Association of North Central Texas.